SVM-based FA / FNA intelligent switching NOB suppression method
Through the SVM classification model and online monitoring data, intelligent switching of FA/FNA inhibition method is realized, solving the automation problem of NOB activity inhibition in short-range nitration processes, improving process stability and reducing costs, and promoting the automated operation of the process.
Patent Information
- Application Number
- CN202411172525.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-26
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2044-08-26
AI Technical Summary
In the short-range nitrification anaerobic ammonia oxidation process, it is difficult to automate the activity inhibition of nitrite oxidation bacteria, resulting in high process stability and cost problems.
The support vector machine (SVM) classification model is adopted, combined with online monitoring data, and the model is trained by optimizing algorithms, intelligent switching of FA/FNA suppression method is realized, and whether to switch the suppression method is automatically judged to ensure long-term stable suppression of NOB.
Based on existing equipment, it is possible to determine whether the water NAR is less than 96% with high accuracy, automatically switch the suppression method, reduce operation costs, improve process stability and efficiency, and promote the automated operation of alternating FA/FNA suppression. .
Smart Images

Figure CN119059644A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to sewage treatment technology, and in particular to a FA / FNA intelligent switching NOB suppression method based on support vector machine (SVM). Background Art
[0002] The short-range nitrification anaerobic ammonium oxidation process (PN / A) is an emerging biological denitrification process in recent years. Compared with traditional biological treatment processes, it can reduce 100% of carbon source addition, 90% of residual sludge production and 60% of aeration energy consumption. It is regarded as an alternative to traditional biological denitrification processes.
[0003] In the short-cut nitrification process (PN), aerobic ammonia oxidizing bacteria (AOB) can oxidize ammonia nitrogen into nitrite nitrogen, and then nitrite nitrogen and ammonia nitrogen act as substrates to participate in the subsequent anaerobic ammonium oxidation reaction. However, in the short-cut nitrification process, the presence of nitrite oxidizing bacteria (NOB) will further oxidize nitrite nitrogen into nitrate nitrogen, thereby reducing the substrate supply required for the anaerobic ammonium oxidation reaction. This will pose a threat to the stable operation of the PN / A process. Therefore, how to stably inhibit the activity of NOB has become the key to the stable operation of the PN / A process.
[0004] In recent years, some studies have shown that alternating inhibition of free ammonia (FA) and free nitrite (FNA) is an effective means to inhibit the growth of NOB. By switching back and forth between the two inhibition methods, NOB can be prevented from adapting to a long-term single inhibition method, thereby inhibiting the activity of NOB for a long time. However, the current FA / FNA alternating inhibition method mainly relies on manual operation, and whether the effluent water quality meets the requirements is determined by manual judgment to decide whether to switch the inhibition method. The judgment criteria for strategy switching are mainly the nitrate nitrogen concentration or nitrite nitrogen accumulation rate (NAR) of the effluent, but the price of online nitrate nitrogen analyzers is expensive, making these judgment criteria difficult to measure online, and thus making it difficult to achieve automated promotion of FA / FNA alternating inhibition. However, in addition to being directly reflected by measuring nitrate nitrogen concentration, these evaluation indicators can also be indirectly reflected by measuring other water quality indicators. These water quality indicators are relatively easy to obtain through conventional online monitoring instruments, such as pH, DO, aeration rate, etc. There is a complex correspondence between these indicators and indicators such as effluent NAR, which is difficult to accurately judge by human experience or traditional control systems.
[0005] In recent years, thanks to its powerful data processing capabilities and variable relationship insight capabilities, the application of machine learning in water quality prediction and process decision-making and early warning has become increasingly common. Among them, support vector machine (SVM) is a machine learning method for classification and regression analysis. Based on statistical learning theory and the principle of maximizing data intervals, sample points of different categories are separated as much as possible. For linearly separable problems, SVM has excellent classification performance and has a wide range of applications in the field of decision-making and early warning, such as whether the effluent water quality is abnormal and whether to perform process operations. Therefore, it is highly feasible to use SVM to determine whether it is necessary to switch the suppression method, which is of practical significance for the automation of FA / FNA alternating suppression. This can not only improve the efficiency and stability of the process, but also further reduce operating costs and promote the promotion of PN / A process in practical applications. Summary of the invention
[0006] In view of the problem that it is difficult to directly measure the switching standard of the inhibition method in FA / FNA alternating inhibition online, the present invention provides a FA / FNA intelligent switching NOB inhibition method based on SVM, trains the SVM classification model with 207 days of reactor operation data, and optimizes the model structure using an optimization algorithm, so that the trained model can accurately determine whether the effluent NAR is lower than 96%, and then decide whether to switch the inhibition method, thereby ensuring the long-term stable inhibition of NOB in the short-range nitrification process, and providing technical support for realizing the automated operation of the FA / FNA alternating inhibition method.
[0007] To achieve the above object, the present invention provides a FA / FNA intelligent switching NOB suppression method based on SVM, comprising the following steps:
[0008] A continuous flow short-range nitrification reactor was built and operated under different FA and FNA concentrations. The calculation formulas for FA and FNA are as follows. Therefore, the pH of the reactor can be controlled by adding an alkali pump, thereby controlling the FA and FNA concentrations of the reactor. The operating parameters of the reactor under different FA and FNA concentrations and the inlet and outlet water quality data were collected for training the SVM classifier.
[0009]
[0010]
[0011] Furthermore, the operating data of the online pH meter, online DO detector, water inlet pump frequency, etc. are transmitted to the client database through the control unit;
[0012] Furthermore, the nitrate nitrogen concentration in the effluent was measured by a spectrophotometer, and the nitrite nitrogen accumulation rate NAR was calculated;
[0013] Furthermore, the collected 207 days of data were annotated according to the NAR obtained in the previous step, and the sample set with NAR higher than 96% was annotated as 0, and the sample set with NAR lower than 96% was annotated as 1;
[0014] Furthermore, aeration volume, aeration coefficient, DO, inlet ammonia nitrogen concentration, ammonia oxidation rate and outlet nitrite nitrogen concentration were selected as the input of the model, the SVM classifier was trained using the classified data in the previous step, and the model was optimized using the grid search cross-validation optimization algorithm to improve the prediction accuracy of the model;
[0015] Furthermore, a control program is written in the control unit so that when the output of the SVM model is 0, the corresponding control program is FNA suppression, and when the output is 1, the corresponding program is FA suppression;
[0016] Furthermore, the trained SVM model is used to make real-time decisions on the FA / FNA alternating suppression method, and the control unit is used to realize data transmission and control of FA and FNA concentrations, thereby achieving intelligent switching of FA / FNA suppression methods.
[0017] Technical Effects
[0018] The present invention proposes a FA / FNA intelligent switching NOB inhibition method based on SVM, in which the operating parameters of the reactor and the water quality data of the inlet and outlet water are input into a trained SVM classification model through a control unit, and the effluent NAR can be judged through conventional online monitoring instrument data, and then a decision is made on whether the inhibition method needs to be switched, thereby achieving long-term effective inhibition of NOB and long-term stable operation of the short-range nitrification process.
[0019] (1) The SVM classification model introduced in the present invention can analyze and learn the relationship between different data variables, use easily measurable water quality data and operating parameters to judge the effluent NAR, and then decide whether to switch the inhibition method. The SVM classifier optimized by grid search cross-validation has an accuracy of 0.90 and an F1 score of 0.92, achieving good prediction results. Compared with the traditional FA / FNA alternating inhibition method that relies on human judgment, the intelligent switching inhibition method proposed in the present invention can predict whether the effluent NAR is lower than 96% without additional equipment such as an online nitrate nitrogen analyzer, and can judge whether the inhibition method needs to be switched, making it possible for sewage treatment plants to achieve FA / FNA alternating inhibition based on existing equipment.
[0020] (2) The control unit of the present invention can automatically execute different inhibition methods according to the output results of the SVM classifier, and realize automatic switching between FA inhibition and FNA inhibition. By presetting the inhibition concentrations of FA and FNA in the control program, the control unit controls the alkali pump to adjust the pH, thereby reaching the predetermined inhibition concentration. At the same time, considering the situation that short-term nitrification consumes alkalinity, the present invention uses FNA inhibition as a long-term inhibition method and FA inhibition as a short-term inhibition method. In this way, the amount of alkali added can be reduced while ensuring the inhibition of NOB, greatly reducing the operating cost.
[0021] In combination with the above advantages, the SVM-based FA / FNA intelligent switching NOB suppression method proposed in the present invention has significant technical effects such as high accuracy, popularization and reduced operating costs. It can effectively solve the problem that the existing FA / FNA alternating suppression method requires human intervention, and can realize the decision-making and automatic switching of the suppression method, providing important technical support for the automated operation of the FA / FNA alternating suppression method. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] Figure 1 It is a schematic diagram of an operating device of a FA / FNA intelligent switching NOB suppression method based on SVM of the present invention.
[0023] Figure 2 It is a real-time operation effect diagram of a preferred embodiment of the FA / FNA intelligent switching NOB suppression method based on SVM of the present invention.
[0024] Figure 3 It is a real-time operation effect diagram of another preferred embodiment of the FA / FNA intelligent switching NOB suppression method based on SVM of the present invention. DETAILED DESCRIPTION
[0025] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
[0026] In the following description, specific details such as specific internal procedures and techniques are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.
[0027] A FA / FNA intelligent switching NOB suppression method based on SVM, the steps are as follows:
[0028] Build a short-range nitrification reactor and its control unit, such as Figure 1 As shown. The bottom of the reactor is equipped with an aeration pipe and a rotor flowmeter is used to control the aeration amount of the air pump into the reactor; the reactor is equipped with an online pH monitor, an online DO monitor and an online ammonia nitrogen monitor to monitor the changes in pH, DO and ammonia nitrogen concentrations; the alkali pump adjusts the pH by adding sodium bicarbonate to the reactor to make FA or FNA reach the set concentration and ensure that the pH is not lower than 6.7; there is a heating rod in the reactor to keep the reaction system at 34°C and keep AOB in suitable environmental conditions.
[0029] The control unit realizes data interaction with the programming platform through the Modbus TCP protocol. The control unit transmits the collected operating parameters and water quality data to the SVM classification model in the programming platform, and then the control unit switches different suppression methods according to the results returned by the model.
[0030] The details are as follows:
[0031] First, a short-range nitrification reactor was built and stably operated for 207 days. During these 207 days, aeration volume, aeration coefficient, DO, influent ammonia nitrogen concentration, ammonia oxidation rate, effluent nitrite nitrogen concentration, NAR and other indicators were collected. Then, the data group was labeled according to whether NAR exceeded 96%. The sample group with NAR lower than 96% was marked as 1, and the sample group with NAR higher than 96% was marked as 0. Then, the SVM classification model was trained with these 207 groups of labeled data. At this time, the model prediction accuracy obtained by training was 0.81 and the F1 score was 0.79. In order to further improve the prediction accuracy of the model and enhance the robustness of the model, the present invention uses grid search to try different model hyperparameters and determine the best model hyperparameter combination through 5-fold cross validation. After optimization, the accuracy of the SVM classification model on the validation set was improved to 0.90, and the F1 score was improved to 0.92, indicating that the accuracy and reliability of the model have been significantly improved.
[0032] The target wastewater treated by the present invention has insufficient alkalinity, and a certain amount of alkalinity will be consumed in the short-range nitrification process. Therefore, in this case, long-term FNA inhibition is easier to achieve, and FA inhibition requires the addition of a certain amount of additional alkalinity to achieve. In order to reduce operating costs, the present invention uses the FNA inhibition method as a long-term inhibition method, that is, the FNA inhibition method is adopted when the output result of the SVM classification model is 0, and the FA inhibition method is adopted when the output result is 1. Then, the corresponding control program is written in the control unit, and the set inhibition concentrations of FA and FNA can be modified through the human-computer interaction interface. The control unit will automatically adjust the pH of the reactor according to the collected effluent ammonia nitrogen concentration and nitrite nitrogen concentration so that FA and FNA reach the set inhibition concentration. At the same time, in order to ensure that the strategy after switching has enough time to inhibit NOB, the present invention sets the time interval for switching between different inhibition methods to 4 days.
[0033] At this time, the trained model and the written control program are used to control the real-time operation of the reactor. The control unit only needs to collect easily available indicators such as aeration volume, aeration coefficient, DO, inlet ammonia nitrogen concentration, ammonia oxidation rate, and effluent nitrite nitrogen concentration to complete the intelligent switching of FA / FNA inhibition methods. The automatic operation of the alternating inhibition method can be achieved without manual operation.
[0034] Example 1
[0035] This embodiment provides a FA / FNA intelligent switching NOB suppression method based on SVM, the reaction device is as follows Figure 1 As shown. In this embodiment, the Python platform is used to realize data interaction with the control unit and the training and application of the SVM model. The data collected by the platform are classified according to whether the effluent NAR is lower than 96%, and the data higher than 96% is marked as 0, and the data lower than 96% is marked as 1. The aeration volume, aeration coefficient, DO, influent ammonia nitrogen concentration, ammonia oxidation rate and effluent nitrite nitrogen concentration are selected as the input of the model to train the SVM classification model, and the model structure is optimized by grid search cross-validation to improve the prediction accuracy of the model. Then the control program is written in the control unit so that when the model output result is 0, it corresponds to the FNA inhibition method, and when the output result is 1, it corresponds to the FA inhibition method. The FA and FNA inhibition concentrations are set in the program, and the set inhibition concentrations can be adjusted at any time through the touch screen and the like. Subsequently, the control unit controls the alkali pump to adjust the pH according to the ammonia nitrogen concentration and nitrite nitrogen concentration in the reactor so that the FA and FNA concentrations in the reactor reach the set inhibition concentrations.
[0036] On the 211th day of the experiment, the reactor's inlet ammonia nitrogen concentration was 1618 mg / L, the outlet nitrite nitrogen concentration was 1043.05 mg / L, the aeration volume was 10 L / min, the aeration coefficient was 99.51 L / g, the DO concentration was 0.25 mg / L, and the ammonia oxidation rate was 83.58%. At this time, the output result of the SVM was 1, that is, the model believed that the FA inhibition method should be switched to at this time, and the control unit automatically increased the pH to perform FA inhibition. To verify the effectiveness of the intelligent switching model, the outlet nitrate nitrogen concentration was measured with a spectrophotometer at this time, and the outlet nitrate nitrogen concentration was 40.38 mg / L, and the NAR was 96.27%. On the 222nd day of the experiment, the reactor's inlet ammonia nitrogen concentration was 1618 mg / L, the outlet nitrite nitrogen concentration was 843.03 mg / L, the aeration volume was 10 L / min, the aeration coefficient was 99.51 L / g, the DO concentration was 0.15 mg / L, the ammonia oxidation rate was 75.36%, and the outlet nitrate nitrogen concentration measured at this time was 40.57 mg / L, and the NAR was 95.41%, which was lower than the training switching standard. Although the SVM classification result was 0 at this time, the program set the switching interval of different strategies to be no less than 4 days, so that FA inhibition was still performed at this time. This embodiment shows that the trained SVM classification model has good robustness and generalization, and can make decision warnings in a timely and accurate manner.
[0037] Example 2
[0038] The apparatus and experimental steps of this embodiment are the same as those of Example 1, and will not be repeated here. On the 226th day of the experiment, the inlet ammonia nitrogen concentration of the reactor was 1600.4 mg / L, the outlet nitrite nitrogen concentration was 722.5 mg / L, the aeration volume was 4.6 L / min, the aeration coefficient was 99.88 L / g, the DO concentration was 0.13 mg / L, the ammonia oxidation rate was 58.82%, the measured outlet nitrate nitrogen concentration was 61.5 mg / L, and the NAR was 92.16%. At this time, the output result of the SVM classifier was 1, and the decision results of the SVM for the next 6 consecutive days were all to adopt the FA inhibition method. On the 231st day of the experiment, the control unit switched the inhibition method back to the FNA inhibition method according to the classification results of the SVM. Possibly affected by the AOB activity at this time, NAR did not reach the set switching standard, but the measured effluent nitrate nitrogen concentration had dropped to 32.07 mg / L, proving that NOB activity had been significantly inhibited, demonstrating the effectiveness of the SVM-based FA / FNA intelligent switching NOB inhibition method of the present invention.
[0039] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. A FA / FNA intelligent switching NOB suppression method based on SVM, characterized in that: The SVM classifier is used to realize the decision-making and early warning of the NOB inhibition method in the continuous flow short-range nitrification reactor, and the automatic switching of the NOB inhibition method in the short-range nitrification process is realized, including the following steps: Collect reactor operation data and inlet and outlet water quality parameters during a continuous flow partial short-range nitrification process; The collected data is transmitted to the client's programming platform through the control unit, and the SVM classification model is trained using the data; The trained SVM model is used to predict the nitrite nitrogen accumulation rate (NAR) of the reactor effluent according to the input real-time operation data of the reactor, and the prediction results are output; The control unit selects and executes the FA or FNA suppression method according to the output result of the SVM model to suppress NOB.
2. The method for suppressing NOB in FA / FNA intelligent switching based on SVM as claimed in claim 1, characterized in that: The method also includes building a short-range nitrification reactor and a supporting control unit, wherein the reactor comprises: Aeration system, used to provide oxygen to the reactor, Online pH monitor, online DO monitor and online ammonia nitrogen monitor are used to monitor the pH, dissolved oxygen concentration and ammonia nitrogen concentration of the reactor in real time, respectively; Alkali pump is used to adjust the pH value to ensure that FA (free ammonia) or FNA (free nitrite) can reach the set inhibition concentration, while keeping the pH in the reactor not lower than 6.7; The heating rod is used to maintain the temperature in the reactor suitable for the growth of AOB (ammonia oxidizing bacteria).
3. A method for suppressing NOB in FA / FNA intelligent switching based on SVM as claimed in claim 1 or 2, characterized in that: The data group is labeled and classified according to whether the NAR exceeds the preset threshold: the data group with NAR below the threshold is marked as 1, indicating that FA suppression should be adopted at this time; the data group with NAR above the threshold is marked as 0, indicating that FNA suppression should be adopted.
4. The method for suppressing NOB in FA / FNA intelligent switching based on SVM as claimed in claim 3, characterized in that: More than 200 days of FA / FNA alternating inhibition operation data were used to train the SVM model for classification, and the model was optimized using grid search and cross validation.
5. The method for suppressing NOB in FA / FNA intelligent switching based on SVM as claimed in claim 1, characterized in that: The control unit can make decisions on the suppression method based on the results returned by the classifier and automatically adjust the operating parameters.
6. A method for suppressing NOB in FA / FNA intelligent switching based on SVM as claimed in claim 5, characterized in that: The trained SVM classification model and the written control program are used to control the real-time operation of the reactor. The control unit only needs to collect easily available indicators such as aeration volume, aeration coefficient, DO, inlet ammonia nitrogen concentration, ammonia oxidation rate and outlet nitrite nitrogen concentration to realize the intelligent switching of FA / FNA inhibition strategy.
Citation Information
Patent Citations
Hybrid intelligent algorithm-based soft measurement method for effluent ammonia nitrogen of anaerobic system
CN108549740A
Method and device for realizing stable low-carbon denitrification of urban sewage by utilizing PN / A
CN115745176A
Method for reducing PN reaction dosage by intelligent control
CN117756268A
Online detection system based on prediction model of nitrite nitrogen in water body
CN214669009U
A method for predicting operation effectiveness of decentralized sewage treatment facility by using support vector machine
US20220316994A1
Cited By
Half short-cut nitrification method for middle-aged and old-aged landfill leachate
CN120328732A
