Intelligent coagulant adding method for waterworks

Abstract

The invention discloses an intelligent coagulant adding method for a waterworks. The method comprises the following steps: acquiring test data and production statistical data of the waterworks in a historical time period; performing data processing on the test data and the production statistical data to obtain a target data set for training; training the initial addition model by using the target data set to obtain a target addition model; obtaining current monitoring data of the water quality of the water plant, and inputting the current monitoring data into the target adding model to obtain a predicted adding amount of the coagulant; on the basis of the factory water turbidity monitored in real time, the feedback regulating quantity of the predicted adding quantity is determined; and performing feedback adjustment on the predicted dosage based on the feedback adjustment quantity to obtain a target dosage corresponding to the current monitoring data, and adding the target dosage to the waterworks. The condition that the outgoing water quality does not reach the standard when the incoming water quality abnormally fluctuates is avoided, and stable water quality control of a water plant is ensured.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for water plants, and in particular to an intelligent dosing method, device and storage medium for coagulants in waterworks. Background Technology

[0002] Water purification processes are crucial to the health and safety of urban residents, with coagulant dosing being a key step in improving water quality. Currently, machine learning is being applied in the field of intelligent chemical dosing at water treatment plants. This technology trains a machine learning model based on historical data from manual dosing, and the predicted dosing amount from the trained model is then added to the water treatment plant, thus completing the coagulant dosing. However, historical data from manual dosing may contain significant errors, leading to low accuracy in the predicted dosing amount by the trained model. Furthermore, the lack of monitoring and adjustment at the water treatment plant after the predicted dosing increases the risk of under- or over-dosing, potentially resulting in substandard water quality and an inability to maintain stable water quality control at the plant. Summary of the Invention

[0003] The present invention aims to at least partially solve one of the technical problems in the related art.

[0004] To address this, the present invention proposes an intelligent coagulant dosing method for water treatment plants. This method involves processing laboratory data and production statistics to obtain a target dataset for training. The initial dosing model is then trained using this target dataset to obtain a target dosing model. After determining the predicted coagulant dosage based on the target dosing model, a feedback adjustment amount is determined based on the real-time monitored turbidity of the effluent. This improves the accuracy of the target dataset, making the predicted dosage of the target dosing model more accurate. Furthermore, the obtained feedback adjustment amount is used for feedback adjustment, preventing substandard effluent quality when influent water quality fluctuates abnormally, thus ensuring stable water quality control at the water treatment plant.

[0005] Another objective of this invention is to provide an intelligent coagulant dosing device for waterworks.

[0006] To achieve the above objectives, the present invention provides a method for intelligent dosing of coagulants in water treatment plants, the method comprising:

[0007] Obtain laboratory and production statistics data from waterworks within a historical time period;

[0008] The test data and production statistics are processed to obtain the target dataset for training.

[0009] The initial dosing model is trained using the target dataset to obtain the target dosing model;

[0010] Obtain the current monitoring data of the water quality of the water plant, input the current monitoring data into the target dosing model, and obtain the predicted dosage of coagulant;

[0011] The feedback adjustment amount of the predicted dosage is determined based on the real-time monitoring of the turbidity of the effluent.

[0012] The predicted dosage is adjusted based on the feedback adjustment amount to obtain the target dosage corresponding to the current monitoring data, and the target dosage is added to the waterworks.

[0013] The intelligent dosing method for coagulants in water treatment plants according to embodiments of the present invention may also have the following additional technical features:

[0014] In one embodiment of the present invention, the step of processing the test data and the production statistics to obtain a target dataset for training includes:

[0015] The test data and the production statistics are combined to obtain the first dataset;

[0016] The first dataset is preprocessed to obtain the second dataset;

[0017] Based on the second dataset, determine the index parameters related to the coagulant dosage;

[0018] The datasets corresponding to the index parameters in the second dataset and the corresponding coagulant dosage are determined as the target datasets for training.

[0019] In one embodiment of the present invention, the preprocessing of the first dataset to obtain the second dataset includes:

[0020] Perform a first missing value processing on the first parameter in the first dataset to obtain the third dataset;

[0021] The second missing value processing is performed on the second parameter in the third dataset to obtain the fourth dataset;

[0022] The third and fourth parameters in the fourth dataset are escaped to obtain the fifth dataset;

[0023] The fifth parameter in the fifth dataset is then de-noiseed and corrected to obtain the second dataset.

[0024] In one embodiment of the present invention, the initial deployment model includes at least one base model and a meta-model; the step of training the initial deployment model using the target dataset to obtain the target deployment model includes:

[0025] Based on the target dataset, a first training dataset and a first test dataset are obtained for training each base model.

[0026] The base models are trained based on the first training dataset to obtain the trained base models and the first prediction dataset of each base model.

[0027] Based on the first test dataset, each trained base model is tested and optimized to obtain each target base model and a second prediction dataset for each base model.

[0028] The meta-model is trained based on the first and second prediction datasets of each of the base models to obtain the target meta-model;

[0029] The target base models and the target meta-models are determined as the target delivery models.

[0030] In one embodiment of the present invention, training the meta-model based on a first prediction dataset of each of the base models and a second prediction dataset of each of the base models to obtain a target meta-model includes:

[0031] The first prediction datasets of each of the base models are stacked to obtain the second training dataset;

[0032] The second prediction datasets of each of the base models are stacked to obtain the second test dataset;

[0033] The meta-model is trained based on the second training dataset to obtain the trained meta-model;

[0034] The trained meta-model is tested and optimized based on the second test dataset to obtain the target meta-model.

[0035] In one embodiment of the present invention, the method further includes:

[0036] Based on the target dataset, at least one base model is determined using model evaluation metrics.

[0037] In one embodiment of the present invention, the base model includes at least one of a support vector machine and a random forest model.

[0038] In one embodiment of the present invention, determining the feedback adjustment amount of the predicted dosage based on the real-time monitored turbidity of the effluent includes:

[0039] Based on real-time monitoring of the turbidity of the treated water, the corresponding feedback adjustment coefficient is determined;

[0040] The product of the predicted dosage and the feedback adjustment coefficient is determined as the feedback adjustment amount of the predicted dosage.

[0041] In one embodiment of the present invention, determining the corresponding feedback adjustment coefficient based on the real-time monitored turbidity of the effluent includes:

[0042] If the turbidity of the effluent is greater than or equal to the standard value, then the corresponding feedback adjustment coefficient is determined to be the first preset value;

[0043] If the turbidity of the effluent is greater than or equal to the upper limit value and less than the standard value, then the corresponding feedback adjustment coefficient is determined to be the second preset value;

[0044] If the turbidity of the effluent is greater than or equal to the ideal value and less than the upper limit value, then the corresponding feedback adjustment coefficient is determined to be the third preset value;

[0045] If the turbidity of the effluent is greater than or equal to the lower limit and less than the ideal value, then the corresponding feedback adjustment coefficient is determined to be the fourth preset value.

[0046] If the turbidity of the effluent is less than the lower limit, then the corresponding feedback adjustment coefficient is determined to be the fifth preset value.

[0047] Another aspect of the present invention proposes an intelligent coagulant dosing device for waterworks, the device comprising:

[0048] The acquisition module is used to acquire test data and production statistics of the waterworks within a historical time period.

[0049] The data processing module is used to process the test data and the production statistics to obtain the target dataset for training.

[0050] The training module is used to train the initial dosing model using the target dataset to obtain the target dosing model;

[0051] The dosing module is used to acquire the current monitoring data of the water quality in the water plant, input the current monitoring data into the target dosing model, and obtain the predicted dosing amount of coagulant;

[0052] The determination module is used to determine the feedback adjustment amount of the predicted dosage based on the real-time monitored turbidity of the effluent.

[0053] The adjustment module is used to perform feedback adjustment on the predicted dosage based on the feedback adjustment amount to obtain the target dosage corresponding to the current monitoring data, and to add the target dosage to the waterworks.

[0054] The intelligent coagulant dosing method and apparatus for water treatment plants according to embodiments of the present invention can process laboratory data and production statistics to obtain a target dataset for training. The target dataset is then used to train an initial dosing model to obtain a target dosing model. After obtaining the predicted dosing amount of coagulant based on the target dosing model, the feedback adjustment amount of the predicted dosing amount is determined based on the real-time monitored turbidity of the effluent. This improves the accuracy of the target dataset, making the predicted dosing amount of the target dosing model more accurate. The obtained feedback adjustment amount is used for feedback adjustment to avoid the situation where the effluent water quality fails to meet the standards when the influent water quality fluctuates abnormally, thus ensuring stable water quality control in the water treatment plant.

[0055] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0056] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0057] Figure 1 This is a flowchart of a method for intelligent dosing of coagulants in a waterworks according to an embodiment of the present invention;

[0058] Figure 2 This is a structural diagram of an intelligent coagulant dosing device for a waterworks according to an embodiment of the present invention. Detailed Implementation

[0059] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0060] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0061] The following description, with reference to the accompanying drawings, describes a method and apparatus for intelligent dosing of coagulants in water treatment plants according to embodiments of the present invention.

[0062] Figure 1 This is a flowchart of the intelligent dosing method for coagulants in water treatment plants according to an embodiment of the present invention.

[0063] like Figure 1 As shown, the method may include the following steps:

[0064] Step 101: Test data and production statistics of the waterworks within a historical time period.

[0065] In one embodiment of the present invention, the aforementioned historical time period can be set as needed, such as 1 year.

[0066] In one embodiment of the present invention, the above-mentioned test data may include sampling date, influent turbidity, influent color, visible matter, influent oxygen consumption, influent pH value, and influent temperature.

[0067] Furthermore, in one embodiment of the present invention, the aforementioned production statistics may include the statistical date and the amount of water entering the plant (unit: m³). 3 ), Water output (unit: m³) 3 ), coagulant dosage (kg), and chemical consumption per thousand tons of water.

[0068] Step 102: Process the test data and production statistics to obtain the target dataset for training.

[0069] In one embodiment of the present invention, after obtaining the test data and production statistics through the above steps, the test data and production statistics can be processed to obtain the target dataset for training.

[0070] In one embodiment of the present invention, the method for processing laboratory data and production statistics to obtain a target dataset for training may include the following steps:

[0071] Step 1021: Merge the test data and production statistics to obtain the first dataset;

[0072] In one embodiment of the present invention, laboratory data and production statistics can be combined empirically to obtain a first dataset. In another embodiment, the data in the first dataset may include the statistical date, influent flow rate (m³ / h), influent turbidity (NTU), effluent turbidity (NTU), color (degrees, common configurations: 5, 10, 15, 20, 25, 30; 99 represents black), visible matter (1: present, 0: absent), influent oxygen consumption (mg / L), effluent oxygen consumption (mg / L), influent ammonia nitrogen (mg / L), effluent ammonia nitrogen (mg / L), influent pH, influent temperature (°C), effluent temperature (°C), oxygen consumption per unit (mg / L), and coagulant dosage (kg).

[0073] Step 1022: Preprocess the first dataset to obtain the second dataset;

[0074] Furthermore, in one embodiment of the present invention, after obtaining the first dataset through the above steps, the first dataset can be preprocessed to make the data in the obtained second dataset continuous and accurate.

[0075] In one embodiment of the present invention, the method for preprocessing the first dataset to obtain the second dataset may include the following steps:

[0076] Step 10221: Perform the first missing value processing on the first parameter in the first dataset to obtain the third dataset;

[0077] Step 10222: Perform second missing value processing on the second parameter in the third dataset to obtain the fourth dataset;

[0078] Step 10223: Escape the third and fourth parameters in the fourth dataset to obtain the fifth dataset;

[0079] Step 10224: Denoise and correct the fifth parameter in the fifth dataset to obtain the second dataset.

[0080] In one embodiment of the present invention, the first parameter may be unit oxygen consumption. Furthermore, the method for performing a first missing value processing on the first parameter in the first dataset to obtain a third dataset may include: performing a first missing value processing on the unit oxygen consumption in the first dataset based on the weekly average value of unit oxygen consumption to obtain the third dataset.

[0081] Furthermore, in one embodiment of the present invention, the second parameter may be the inlet water temperature. Also, the method for performing a second missing value processing on the second parameter in the third dataset to obtain the fourth dataset may include: using linear regression prediction based on the outlet water temperature to obtain and supplement the data in the third dataset where the inlet water temperature is missing.

[0082] Furthermore, in one embodiment of the present invention, the third parameter can be chroma, and the fourth parameter can be visible matter. Also, in one embodiment of the present invention, the method for escaping the third and fourth parameters in the fourth dataset to obtain the fifth dataset may include: chroma is assigned a value during the period, with a minimum value of 5, and is escaped to 5, 10, 15, 20, 25, and 30 respectively according to the upward alignment method; values ​​greater than 30 are taken as 99 (representing black); visible matter is escaped using 0 and 1, with visible matter or floating matter being escaped as 1, otherwise escaped as 0 (nothing).

[0083] Furthermore, in one embodiment of the present invention, the fifth parameter may be the coagulant dosage. Also, in one embodiment of the present invention, the method for denoising and correcting the fifth parameter in the fifth dataset to obtain the second dataset may include: denoising the noisy data of the coagulant dosage in the fifth dataset, and correcting the denoised coagulant dosage using a correction function to obtain the second dataset.

[0084] In one embodiment of the present invention, noise data in the coagulant dosage can be determined by the standard deviation, and the noise data in the coagulant dosage can be deleted from the fifth dataset, thus completing the noise reduction processing of the coagulant dosage in the fifth dataset. In one embodiment of the present invention, the method for determining noise data in the coagulant dosage by the standard deviation may include: calculating the average value (μ) and standard deviation (σ) of the coagulant dosage, setting a reasonable range for the coagulant [μ-10σ, μ+10σ] based on the average value and standard deviation, and determining the coagulant dosage data that exceeds the reasonable range as noise data in the coagulant dosage.

[0085] Furthermore, in one embodiment of the present invention, the above-mentioned correction function is:

[0086]

[0087] Where x is the turbidity of the treated water, y is the dosage correction coefficient for the coagulant dosage, and b is the target variation range parameter.

[0088] It should be noted that, in one embodiment of the present invention, the parameter range corresponding to the variation amplitude parameter b can be traversed, and the denoised coagulant dosage can be corrected using the correction function corresponding to the variation amplitude parameter b. Then, the initial dosing model is trained using the corrected coagulant dosage to determine the evaluation effect of the trained dosing model. The variation amplitude parameter with the optimal evaluation effect of the trained dosing model is determined as the target variation amplitude parameter. In one embodiment of the present invention, the parameter range corresponding to the variation amplitude parameter b can be set as needed, such as [2, 10]. And, in one embodiment of the present invention, the R-value of the trained dosing model can be determined. 2 (Coefficient of determination), MSE (mean squared error), and MAE (mean absolute error), and R 2 The average values ​​of MSE and MAE are used to evaluate the performance of the trained loading model.

[0089] Step 1023: Based on the second dataset, determine the index parameters related to the coagulant dosage;

[0090] In one embodiment of the present invention, the Person method can be used to analyze the correlation of parameters in the second dataset to obtain index parameters related to the coagulant dosage. These index parameters may include date, influent flow rate, influent turbidity, influent pH, color, visible matter, temperature, and gas consumption.

[0091] Step 1024: Determine the datasets corresponding to the index parameters in the second dataset and the corresponding coagulant dosage as the target dataset for training.

[0092] Step 103: Train the initial dosing model using the target dataset to obtain the target dosing model.

[0093] In one embodiment of the present invention, after determining the target dataset through the above steps, the initial dosing model can be trained using the target dataset to obtain the target dosing model, so that the predicted dosing amount of coagulant can be obtained subsequently based on the target dosing model.

[0094] In one embodiment of the present invention, the initial deployment model may include at least one base model and a meta-model. Furthermore, in another embodiment of the present invention, at least one base model may be determined based on the target dataset using model evaluation metrics.

[0095] Specifically, in one embodiment of the present invention, multiple candidate models can be compared based on the sklearn machine learning library, and the candidate models can be evaluated according to the evaluation metrics MSE, MAE, and R. 2 The analysis identifies at least one base model. The candidate models may include linear regression, ridge regression, XGBoost regression, support vector machine, and random forest regression algorithms. Furthermore, the base model may include at least one of support vector machine and random forest models.

[0096] Furthermore, in one embodiment of the present invention, the method for training the initial dosing model using the target dataset to obtain the target dosing model may include the following steps:

[0097] Step 1031: Based on the target dataset, obtain the first training dataset and the first test dataset for training each base model;

[0098] In one embodiment of the present invention, the data in the target dataset can be divided into a first training dataset and a first test dataset for training various base models according to a preset ratio. For example, assuming the preset ratio is 8:2, that is, 80% of the data in the target dataset is randomly divided into the first training dataset and 20% into the first test dataset.

[0099] Step 1032: Train each base model based on the first training dataset to obtain each trained base model and the first prediction dataset of each base model;

[0100] In one embodiment of the present invention, each base model can be trained based on data in the first training dataset to obtain the corresponding predicted coagulant dosage. In one embodiment of the present invention, the training method for each base model is the same as that in the prior art, and will not be described in detail here.

[0101] Step 1033: Test and fine-tune each trained base model based on the first test dataset to obtain each target base model and the second prediction dataset of each base model;

[0102] In one embodiment of the present invention, after obtaining each trained base model through the above steps, the first test dataset can be used to test and fine-tune each trained base model, thereby obtaining the accuracy performance of each trained base model on the first test dataset and evaluating the generalization ability of each trained base model. Then, based on the goodness of fit (R²) of the first test dataset... 2 The mean squared error (MSE) and mean absolute error (MAE) are used as evaluation metrics to adjust and optimize each trained base model, resulting in the second prediction dataset corresponding to each target base model and the first test dataset.

[0103] Step 1034: Train the meta-model based on the first and second prediction datasets of each base model to obtain the target meta-model;

[0104] In one embodiment of the present invention, after obtaining the first prediction dataset and the second prediction dataset of each base model through the above steps, the meta-model can be trained based on the first prediction dataset and the second prediction dataset of each base model to obtain the target meta-model.

[0105] Specifically, in one embodiment of the present invention, the method for training the meta-model based on the first prediction dataset and the second prediction dataset of each base model to obtain the target meta-model may include the following steps:

[0106] Step 10341: Stack the first prediction datasets of each base model to obtain the second training dataset;

[0107] Step 10342: Stack the second prediction datasets of each base model to obtain the second test dataset;

[0108] Step 10343: Train the meta-model based on the second training dataset to obtain the trained meta-model;

[0109] Step 10344: Test and fine-tune the trained meta-model based on the second test dataset to obtain the target meta-model.

[0110] In one embodiment of the present invention, the second training dataset includes the predicted coagulant dosage and the corresponding actual coagulant dosage from the first prediction dataset of each base model. In another embodiment of the present invention, the second test dataset includes the predicted coagulant dosage and the corresponding actual coagulant dosage from the second prediction dataset of each base model.

[0111] Furthermore, in one embodiment of the present invention, after obtaining the second training dataset through the above steps, the meta-model can be trained based on the second training dataset to obtain the trained meta-model, and the trained meta-model can be tested and optimized based on the second test dataset to obtain the target meta-model.

[0112] Furthermore, in one embodiment of the present invention, the method of training the meta-model based on the second training dataset and testing and optimizing the trained meta-model based on the second test dataset can be referred to the detailed description in the above embodiments, and will not be repeated here.

[0113] Step 1035: Determine each target base model and target meta-model as the target delivery model.

[0114] Step 104: Obtain the current monitoring data of the water quality of the water plant, input the current monitoring data into the target dosing model, and obtain the predicted dosage of coagulant.

[0115] In one embodiment of the present invention, after determining the target dosing model through the above steps, the current monitoring data can be obtained after a preset time since the last dosing to the waterworks, and the predicted dosing amount of coagulant can be obtained based on the current monitoring data through the target dosing model.

[0116] In one embodiment of the present invention, the above-mentioned preset time can be set as needed, such as 30 minutes. That is, the current monitoring data can be obtained 30 minutes after the last addition to the waterworks, and the predicted addition amount of coagulant can be obtained based on the current monitoring data through the target addition model.

[0117] In one embodiment of the present invention, current monitoring data of water quality in the water plant can be obtained, and the date, influent flow rate, influent turbidity, influent pH value, color, visible matter, temperature and gas consumption in the current monitoring data can be input into each target basis model to obtain the predicted dosage output by each target basis model. The predicted dosage output by each target basis model can then be input into the target meta-model to obtain the predicted dosage of coagulant.

[0118] It should be noted that, in one embodiment of the present invention, when the system is started for the first time, the predicted dosage can be determined as the target dosage.

[0119] Step 105: Based on the real-time monitoring of the turbidity of the effluent, determine the feedback adjustment amount for the predicted dosage.

[0120] In one embodiment of the present invention, after obtaining the predicted dosage through the above steps, the feedback adjustment amount of the predicted dosage can be determined based on the real-time monitored turbidity of the effluent to complete the feedback adjustment.

[0121] Specifically, in one embodiment of the present invention, the method for determining the feedback adjustment amount of the predicted dosage based on real-time monitoring of the turbidity of the effluent may include the following steps:

[0122] Step 1051: Determine the corresponding feedback adjustment coefficient based on the real-time monitored turbidity of the effluent;

[0123] Step 1052: The product of the predicted dosage and the feedback adjustment coefficient is determined as the feedback adjustment amount of the predicted dosage.

[0124] In one embodiment of the present invention, the method for determining the corresponding feedback adjustment coefficient based on real-time monitoring of the turbidity of the treated water may include the following steps:

[0125] Step 10511: If the turbidity of the effluent is greater than or equal to the standard value, then determine the corresponding feedback adjustment coefficient as the first preset value;

[0126] Step 10512: If the turbidity of the effluent is greater than or equal to the upper limit value and less than the standard value, then the corresponding feedback adjustment coefficient is determined to be the second preset value.

[0127] Step 10513: If the turbidity of the effluent is greater than or equal to the ideal value and less than the upper limit value, then the corresponding feedback adjustment coefficient is determined to be the third preset value.

[0128] Step 10514: If the turbidity of the effluent is greater than or equal to the lower limit and less than the ideal value, then the corresponding feedback adjustment coefficient is determined to be the fourth preset value.

[0129] Step 10515: If the turbidity of the effluent is less than the lower limit, then the corresponding feedback adjustment coefficient is determined to be the fifth preset value.

[0130] In one embodiment of the present invention, the aforementioned standard value, ideal value, upper limit value, and lower limit value can be determined through data experience. In one embodiment of the present invention, the aforementioned standard value can be 1 NTU; the ideal value can be 0.55 ± 0.15 NTU; the upper limit value is 0.8 NTU; and the lower limit value is 0.2 NTU.

[0131] Furthermore, in one embodiment of the present invention, the first preset value, second preset value, third preset value, fourth preset value, and fifth preset value can be determined empirically. In one embodiment of the present invention, the first preset value is +50%, the second preset value is +30%, the third preset value is +10%, the fourth preset value is -10%, and the fifth preset value is -30%.

[0132] In one embodiment of the present invention, if the turbidity of the effluent is greater than or equal to 1 NTU, the corresponding feedback adjustment factor is determined to be +50%; if the turbidity of the effluent is greater than or equal to 0.8 NTU and less than 1 NTU, the corresponding feedback adjustment factor is determined to be +30%; if the turbidity of the effluent is greater than or equal to 0.7 NTU and less than 0.8 NTU, the corresponding feedback adjustment factor is determined to be +10%; if the turbidity of the effluent is greater than or equal to 0.2 NTU and less than 0.4 NTU, the corresponding feedback adjustment factor is determined to be -10%; and if the turbidity of the effluent is less than 0.2 NTU, the corresponding feedback adjustment factor is determined to be -30%.

[0133] In one embodiment of the present invention, after obtaining the feedback adjustment coefficient through the above steps, the product of the predicted dosage and the feedback adjustment coefficient can be determined as the feedback adjustment amount of the predicted dosage, that is, the feedback adjustment amount = predicted dosage × feedback adjustment coefficient.

[0134] Step 106: Adjust the predicted dosage based on the feedback adjustment amount to obtain the target dosage corresponding to the current monitoring data, and add the target dosage to the waterworks.

[0135] In one embodiment of the present invention, after determining the feedback adjustment amount through the above steps, the predicted dosage can be adjusted based on the feedback adjustment amount to obtain the target dosage corresponding to the current monitoring data.

[0136] Furthermore, in one embodiment of the present invention, the method for adjusting the predicted dosage based on the feedback adjustment amount to obtain the target dosage corresponding to the current monitoring data may include: adding the predicted dosage amount to the feedback adjustment amount to obtain the coagulant dosage amount at the next moment, until the predicted dosage amount is determined by the target dosage model next time, and determining the last adjusted coagulant dosage amount as the target coagulant dosage amount corresponding to the current monitoring data.

[0137] For example, in one embodiment of the present invention, it is assumed that at each precise moment (e.g., 10:00), the predicted dosage is determined by the target dosage model, and the corresponding target dosage is obtained based on the determined feedback adjustment amount. Every 30 minutes, the current adjusted coagulant dosage is obtained by using the previous dosage as a benchmark and the corresponding feedback adjustment amount. That is, the current adjusted coagulant dosage = the previous dosage + the feedback adjustment amount. The above steps are repeated until the predicted dosage is determined by the target dosage model at the next moment.

[0138] In one embodiment of the present invention, the above method may include: storing the current monitoring data and the corresponding target coagulant dosage as update data in a database; updating the target dosage model using the update data to obtain the updated dosage model; and using the updated dosage model as the target dosage model.

[0139] The intelligent coagulant dosing method for waterworks according to embodiments of the present invention involves: acquiring laboratory data and production statistics of the waterworks over a historical period; processing the laboratory data and production statistics to obtain a target dataset for training; training an initial dosing model using the target dataset to obtain a target dosing model; acquiring current monitoring data of the water quality at the waterworks and inputting the current monitoring data into the target dosing model to obtain the predicted dosing amount of coagulant; determining a feedback adjustment amount for the predicted dosing amount based on the real-time monitored turbidity of the effluent; adjusting the predicted dosing amount based on the feedback adjustment amount to obtain the target dosing amount corresponding to the current monitoring data, and then adding the target dosing amount to the waterworks. This invention can process laboratory data and production statistics to obtain a target dataset for training. The initial dosing model is then trained using the target dataset to obtain a target dosing model. After obtaining the predicted dosage of coagulant based on the target dosing model, the feedback adjustment amount of the predicted dosage is determined based on the real-time monitored turbidity of the effluent. This improves the accuracy of the target dataset, making the predicted dosage of the target dosing model more accurate. The obtained feedback adjustment amount is used for feedback adjustment to avoid situations where the effluent water quality fails to meet standards when the influent water quality fluctuates abnormally, thus ensuring stable water quality control in the water plant.

[0140] Figure 2 This is a schematic diagram of the structure of the intelligent coagulant dosing device 10 for waterworks according to an embodiment of the present invention.

[0141] like Figure 2 As shown, the device may include:

[0142] Module 201 is used to acquire laboratory data and production statistics of the waterworks within a historical time period.

[0143] Data processing module 202 is used to process laboratory data and production statistics to obtain a target dataset for training.

[0144] Training module 203 is used to train the initial dosing model using the target dataset to obtain the target dosing model;

[0145] The dosing module 204 is used to acquire the current monitoring data of the water quality of the water plant, input the current monitoring data into the target dosing model, and obtain the predicted dosing amount of coagulant;

[0146] The determination module 202 is used to determine the feedback adjustment amount of the predicted dosage based on the real-time monitoring of the turbidity of the effluent.

[0147] The adjustment module 206 is used to perform feedback adjustment on the predicted dosage based on the feedback adjustment amount, obtain the target dosage corresponding to the current monitoring data, and add the target dosage to the waterworks.

[0148] In one embodiment of this disclosure, the data processing module 202 is specifically used for:

[0149] The laboratory data and production statistics were merged to obtain the first dataset;

[0150] Preprocess the first dataset to obtain the second dataset;

[0151] Based on the second dataset, determine the index parameters related to the coagulant dosage.

[0152] The datasets corresponding to the index parameters in the second dataset and the corresponding coagulant dosage are determined as the target datasets for training.

[0153] In one embodiment of this disclosure, the data processing module 202 is further configured to:

[0154] Perform the first missing value processing on the first parameter in the first dataset to obtain the third dataset;

[0155] The second missing value processing is performed on the second parameter in the third dataset to obtain the fourth dataset;

[0156] The third and fourth parameters in the fourth dataset are escaped to obtain the fifth dataset.

[0157] The fifth parameter in the fifth dataset is de-noiseed and corrected to obtain the second dataset.

[0158] In one embodiment of this disclosure, the initial feeding model includes at least one base model and a meta-model; the training module 203 is specifically used for:

[0159] Based on the target dataset, the first training dataset and the first test dataset are obtained for training each base model.

[0160] Each base model is trained based on the first training dataset to obtain each trained base model and the first prediction dataset of each base model;

[0161] Based on the first test dataset, each trained base model is tested and optimized to obtain each target base model and the second prediction dataset of each base model.

[0162] The meta-model is trained based on the first and second prediction datasets of each base model to obtain the target meta-model;

[0163] Each target base model and target meta-model is determined as the target delivery model.

[0164] In one embodiment of this disclosure, the training module 203 is further configured to:

[0165] The first prediction datasets of each base model are stacked to obtain the second training dataset;

[0166] The second prediction datasets of each base model are stacked to obtain the second test dataset;

[0167] The meta-model is trained based on the second training dataset to obtain the trained meta-model;

[0168] The trained meta-model was tested and optimized based on the second test dataset to obtain the target meta-model.

[0169] In one embodiment of this disclosure, the above-described apparatus is further used for:

[0170] Based on the target dataset, at least one base model is determined using model evaluation metrics.

[0171] In one embodiment of this disclosure, the base model includes at least one of support vector machine and random forest model.

[0172] In one embodiment of this disclosure, the determining module 205 is specifically used for:

[0173] Based on real-time monitoring of the turbidity of the treated water, the corresponding feedback adjustment coefficient is determined;

[0174] The product of the predicted dosage and the feedback adjustment coefficient is determined as the feedback adjustment amount of the predicted dosage.

[0175] In one embodiment of this disclosure, the determining module 205 is further configured to:

[0176] If the turbidity of the treated water is greater than or equal to the standard value, then the corresponding feedback adjustment coefficient is determined to be the first preset value.

[0177] If the turbidity of the treated water is greater than or equal to the upper limit but less than the standard value, then the corresponding feedback adjustment coefficient is determined to be the second preset value.

[0178] If the turbidity of the treated water is greater than or equal to the ideal value but less than the upper limit value, then the corresponding feedback adjustment coefficient is determined to be the third preset value.

[0179] If the turbidity of the treated water is greater than or equal to the lower limit but less than the ideal value, then the corresponding feedback adjustment coefficient is determined to be the fourth preset value.

[0180] If the turbidity of the treated water is less than the lower limit, then the corresponding feedback adjustment coefficient is set to the fifth preset value.

[0181] The intelligent coagulant dosing device for waterworks according to this invention acquires laboratory data and production statistics of the waterworks over a historical period; processes the laboratory data and production statistics to obtain a target dataset for training; trains an initial dosing model using the target dataset to obtain a target dosing model; acquires current monitoring data of the water quality of the waterworks and inputs the current monitoring data into the target dosing model to obtain the predicted dosing amount of coagulant; determines the feedback adjustment amount of the predicted dosing amount based on the real-time monitored turbidity of the effluent; adjusts the predicted dosing amount based on the feedback adjustment amount to obtain the target dosing amount corresponding to the current monitoring data, and adds the target dosing amount to the waterworks. This invention can process laboratory data and production statistics to obtain a target dataset for training. The initial dosing model is then trained using the target dataset to obtain a target dosing model. After obtaining the predicted dosage of coagulant based on the target dosing model, the feedback adjustment amount of the predicted dosage is determined based on the real-time monitored turbidity of the effluent. This improves the accuracy of the target dataset, making the predicted dosage of the target dosing model more accurate. The obtained feedback adjustment amount is used for feedback adjustment to avoid situations where the effluent water quality fails to meet standards when the influent water quality fluctuates abnormally, thus ensuring stable water quality control in the water plant.

[0182] In this specification, the use of terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refers to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0183] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A method for intelligent dosing of coagulants in a waterworks, characterized in that, The method includes: Obtain laboratory and production statistics data from waterworks within a historical time period; The test data and production statistics are processed to obtain the target dataset for training. The initial dosing model is trained using the target dataset to obtain the target dosing model; Obtain the current monitoring data of the water quality of the water plant, input the current monitoring data into the target dosing model, and obtain the predicted dosage of coagulant; The feedback adjustment amount of the predicted dosage is determined based on the real-time monitoring of the turbidity of the effluent. The predicted dosage is adjusted based on the feedback adjustment amount to obtain the target dosage corresponding to the current monitoring data, and the target dosage is added to the waterworks.

2. The method according to claim 1, characterized in that, The process of processing the test data and the production statistics to obtain the target dataset for training includes: The test data and the production statistics are combined to obtain the first dataset; The first dataset is preprocessed to obtain the second dataset; Based on the second dataset, determine the index parameters related to the coagulant dosage; The datasets corresponding to the index parameters in the second dataset and the corresponding coagulant dosage are determined as the target datasets for training.

3. The method according to claim 2, characterized in that, The preprocessing of the first dataset to obtain the second dataset includes: Perform a first missing value processing on the first parameter in the first dataset to obtain the third dataset; The second missing value processing is performed on the second parameter in the third dataset to obtain the fourth dataset; The third and fourth parameters in the fourth dataset are escaped to obtain the fifth dataset; The fifth parameter in the fifth dataset is then de-noiseed and corrected to obtain the second dataset.

4. The method according to claim 1, characterized in that, The initial dosing model includes at least one base model and a meta-model; training the initial dosing model using the target dataset to obtain the target dosing model includes: Based on the target dataset, a first training dataset and a first test dataset are obtained for training each base model. The base models are trained based on the first training dataset to obtain the trained base models and the first prediction dataset of each base model. Based on the first test dataset, each trained base model is tested and optimized to obtain each target base model and a second prediction dataset for each base model. The meta-model is trained based on the first and second prediction datasets of each of the base models to obtain the target meta-model; The target base models and the target meta-models are determined as the target delivery models.

5. The method according to claim 4, characterized in that, The meta-model is trained based on the first and second prediction datasets of each of the base models to obtain the target meta-model, including: The first prediction datasets of each of the base models are stacked to obtain the second training dataset; The second prediction datasets of each of the base models are stacked to obtain the second test dataset; The meta-model is trained based on the second training dataset to obtain the trained meta-model; The trained meta-model is tested and optimized based on the second test dataset to obtain the target meta-model.

6. The method according to claim 4, characterized in that, The method further includes: Based on the target dataset, at least one base model is determined using model evaluation metrics.

7. The method according to claim 6, characterized in that, The base model includes at least one of support vector machine and random forest models.

8. The method according to claim 1, characterized in that, The feedback adjustment amount for determining the predicted dosage based on real-time monitored turbidity of the treated water includes: Based on real-time monitoring of the turbidity of the treated water, the corresponding feedback adjustment coefficient is determined; The product of the predicted dosage and the feedback adjustment coefficient is determined as the feedback adjustment amount of the predicted dosage.

9. The method according to claim 8, characterized in that, The determination of the corresponding feedback adjustment coefficient based on the real-time monitoring of the turbidity of the treated water includes: If the turbidity of the effluent is greater than or equal to the standard value, then the corresponding feedback adjustment coefficient is determined to be the first preset value; If the turbidity of the effluent is greater than or equal to the upper limit value and less than the standard value, then the corresponding feedback adjustment coefficient is determined to be the second preset value; If the turbidity of the effluent is greater than or equal to the ideal value and less than the upper limit value, then the corresponding feedback adjustment coefficient is determined to be the third preset value; If the turbidity of the effluent is greater than or equal to the lower limit and less than the ideal value, then the corresponding feedback adjustment coefficient is determined to be the fourth preset value. If the turbidity of the effluent is less than the lower limit, then the corresponding feedback adjustment coefficient is determined to be the fifth preset value.

10. A smart coagulant dosing device for a waterworks, characterized in that, The device includes: The acquisition module is used to acquire test data and production statistics of the waterworks within a historical time period. The data processing module is used to process the test data and the production statistics to obtain the target dataset for training. The training module is used to train the initial dosing model using the target dataset to obtain the target dosing model; The dosing module is used to acquire the current monitoring data of the water quality in the water plant, input the current monitoring data into the target dosing model, and obtain the predicted dosing amount of coagulant; The determination module is used to determine the feedback adjustment amount of the predicted dosage based on the real-time monitored turbidity of the effluent. The adjustment module is used to perform feedback adjustment on the predicted dosage based on the feedback adjustment amount to obtain the target dosage corresponding to the current monitoring data, and to add the target dosage to the waterworks.

Images