AI intelligent aeration method based on sewage treatment

Abstract

The application discloses an AI intelligent aeration method based on sewage treatment and belongs to the field of sewage treatment, aiming to solve the technical problems of traditional aeration, such as end feedback lag, rigid regulation strategy, energy consumption redundancy caused by the lack of standard working condition experience reuse system, and low water quality standard efficiency; the method first deploys a self-cleaning rapid membrane separation device before the inlet of a secondary sedimentation tank, quantifies the membrane pollution state and triggers online backwashing, and synchronously collects the core water quality and biochemical indicators of the non-interference water sample; then, based on the national standard, dynamic target values are set, and through the fan bidirectional regulation logic and coarse and fine regulation strategies, precise aeration regulation and target value iterative optimization are realized; finally, the aeration data is constructed into a working condition template, the new cycle reuses the high-quality parameters through similarity matching, and the template library is clustered and purified; the application realizes front-end precise monitoring, aeration dynamic adaptation and experience closed-loop reuse, guarantees the water quality of effluent to meet the standard, greatly reduces the aeration energy consumption, and improves the intelligent and lean level of the aeration link.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of sewage treatment, and particularly relates to an AI intelligent aeration method based on sewage treatment. BACKGROUND

[0002] The aeration link is the core process of the biochemical reaction stage of sewage treatment, and the aeration intensity directly determines the degradation efficiency of organic pollutants and nitrogen pollutants in sewage. At the same time, it is also the main source of energy consumption of the sewage treatment plant. With the tightening of environmental protection standards and the increasing demand for energy saving and consumption reduction, the limitations of the existing sewage aeration control technology are increasingly prominent. Not only are there problems of signal feedback lag and rigid regulation strategy, but also there is a lack of effective reuse and optimization mechanism for historical regulation experience, resulting in that the aeration process is difficult to guarantee stable and standard water quality, and a large amount of energy consumption redundancy is caused. Specifically, the following three technical problems are embodied:

[0003] The existing sewage treatment plant generally carries out "ex post" feedback aeration according to the end water concentration of the secondary sedimentation tank. The signal naturally lags, and the sensor is easily disturbed by the adhesion of active sludge flocs and suspended particles, resulting in "late measurement and inaccurate measurement". In order to ensure standard, a high aeration allowance must be maintained, resulting in energy consumption redundancy;

[0004] The traditional aeration control takes the national standard discharge value as the fixed upper limit, and the target value is static and single, which cannot be self-adapted to relax or tighten with the changes of influent load and microbial state. The fan regulation only has "turning on to increase oxygen" logic, and lacks "active reduction" mechanism, resulting in over-aeration and high power consumption all the year round;

[0005] There is no standardized working condition experience retention and reuse system for the existing aeration regulation. The core regulation data of single aeration does not form a structured template, and the parameters of the new aeration cycle need to be debugged from zero, which not only prolongs the time for water quality to reach the standard, but also easily causes secondary energy consumption waste due to blind parameter setting. At the same time, a large amount of low-quality and abnormal working condition data is mixed in the long-term accumulated regulation data, and there is no targeted template purification mechanism, which not only occupies the database storage resources, but also interferes with the parameter adaptation under the new working condition, resulting in that the precision and adaptation efficiency of aeration regulation are difficult to continuously improve. Therefore, we propose an AI intelligent aeration method based on sewage treatment. SUMMARY

[0006] The purpose of the present application is to provide an AI intelligent aeration method based on sewage treatment to solve the problems raised in the background art.

[0007] To achieve the above purpose, the present application provides the following technical scheme: an AI intelligent aeration method based on sewage treatment, comprising the following steps:

[0008] Step one: Deploy a rapid membrane separation device before the secondary sedimentation tank inlet, analyze the membrane pollution characterization value of the membrane module, construct a membrane pollution trend change curve according to the value, calculate the membrane pollution early warning evaluation value, judge whether to trigger online backwashing, collect non-interference water samples, and obtain the core water quality indicators of non-interference water samples;

[0009] Step two: Obtain the measured value of the core water quality indicators of non-interference water samples, calculate the dynamic target value of each water quality indicator; take the dynamic target value as the initial control target, adjust the aeration quantity through the fan air volume adjustment mechanism and monitor the measured value of the water quality indicators, calculate the water quality indicator difference value; judge whether to trigger dynamic target value optimization and iteration; determine the fan adjustment direction and coarse adjustment and fine adjustment rules until the water quality indicator deviation meets the accuracy requirement;

[0010] Step three: Integrate the aeration adjustment direction and the corresponding stage correction step to generate the fan speed or frequency command and execute the aeration; store the aeration adjustment full-quantity core data, construct the aeration adjustment working condition comparison template; in a new aeration period, select the optimal adaptive working condition template, extract its initial proportion coefficient and coarse adjustment and fine adjustment step as the initial parameters corresponding to the new working condition; perform low-quality template purification operation on the templates in the database.

[0011] Preferably, the specific process of analyzing the membrane pollution characterization value of the membrane module is as follows:

[0012] A self-cleaning small rapid membrane separation device is deployed before the secondary sedimentation tank inlet. The device is built-in with an online backwashing unit. The surface of the membrane module corresponding to the rapid membrane separation device is uniformly provided with a plurality of detection points along the sewage flow direction. Each detection point is provided with a transmembrane pressure difference sensor.

[0013] A collection period is set. The transmembrane pressure difference value of the membrane surface of each detection point is obtained through the transmembrane pressure difference sensor and recorded as the detection point membrane pollution characterization value. The detection points are sorted according to the sewage flow direction and a basic weight is preset. The basic weight is then normalized to obtain the pollution influence weight of each detection point.

[0014] The membrane pollution characterization value of the membrane module corresponding to the rapid membrane separation device is calculated by combining the membrane pollution characterization value of each detection point and the corresponding pollution influence weight.

[0015] Preferably, the specific process of calculating the membrane pollution early warning evaluation value and judging whether to trigger online backwashing is as follows:

[0016] A rectangular coordinate system is constructed with time as the horizontal coordinate and the membrane pollution characterization value of the membrane module as the vertical coordinate. The membrane pollution characterization value of the membrane module at each collection time is marked based on the initial time of the last self-cleaning of the rapid membrane separation device to obtain pollution data points. The pollution data points are connected in time sequence by a smooth curve to form a membrane pollution trend change curve.

[0017] The pollution increasing rate per unit time is obtained by subtracting the membrane pollution characterization value of the membrane module corresponding to the initial time from the membrane pollution characterization value of the membrane module corresponding to the current acquisition time, and then dividing the total duration of the continuous acquisition period;

[0018] The maximum mutation amplitude of pollution is taken as the maximum increment per unit time of the membrane pollution characterization value of the membrane module from the initial time to the current time;

[0019] The membrane pollution early warning evaluation value is obtained by weighted analysis of the membrane pollution characterization value of the membrane module corresponding to the current acquisition time, the pollution increasing rate per unit time, and the maximum mutation amplitude of pollution;

[0020] If the membrane pollution early warning evaluation value is greater than or equal to the corresponding preset threshold value, the pollution early warning is triggered, and the online backwashing unit is started to perform low-pressure backwashing.

[0021] Preferably, the specific process of collecting the non-interference water sample and obtaining the core water quality indicators and biochemical reaction state indicators of the non-interference water sample is as follows:

[0022] The non-interference water sample is collected through the clean water collection channel of the self-cleaning small rapid membrane separation equipment according to the preset sampling period, and the core water quality indicators and biochemical reaction state indicators of the non-interference water sample are collected;

[0023] The water quality indicators include chemical oxygen demand, ammonia nitrogen, and total nitrogen.

[0024] Preferably, the specific process of monitoring the measured values of the water quality indicators and calculating the difference values of the water quality indicators is as follows:

[0025] The measured values of the core water quality indicators corresponding to the non-interference water sample are obtained;

[0026] The national discharge standard values of each core water quality indicator are obtained;

[0027] A preset initial proportion coefficient is obtained, and the dynamic target values of the core water quality indicators are obtained by multiplying the national discharge standard values of the core water quality indicators by the preset initial proportion coefficient;

[0028] The dynamic target values are taken as the initial control targets, the wastewater to be treated in the biochemical tank is continuously aerated and adjusted through the fan air volume adjustment mechanism, and the measured values of the water quality indicators in the non-interference water sample collected by the rapid membrane separation equipment in each preset sampling period are monitored in real time;

[0029] For each sampling period, the difference values of the core water quality indicators are obtained by subtracting the corresponding dynamic target values from the measured values of the core water quality indicators.

[0030] Preferably, the specific process of judging whether to trigger dynamic target value optimization and iteration is as follows:

[0031] The preset precision threshold, the proportional coefficient increment, the proportional coefficient final value, and the continuous qualified period number;

[0032] If the absolute values of the differences of the three core water quality indexes in the current sampling period are all less than or equal to the preset precision threshold, and the continuous period number remains in this state, the dynamic target value optimization is triggered.

[0033] The specific process of the dynamic target value optimization is as follows:

[0034] The updated proportional coefficient is obtained by adding the proportional coefficient increment to the current proportional coefficient, and the dynamic target value is updated based on the new coefficient.

[0035] The proportional coefficient is repeatedly updated by the proportional coefficient increment, the new dynamic target value is calculated, and it is monitored whether the water quality deviation in each collection period meets the condition that the absolute values of the differences of the three water quality indexes in the continuous period number are all less than or equal to the preset precision threshold, so as to trigger the next optimization cycle, until the proportional coefficient reaches the preset proportional coefficient final value, the dynamic target value iteration is stopped, and the current dynamic target value is kept stable.

[0036] Preferably, the specific process of the fan air volume regulation mechanism is as follows:

[0037] The preset coarse adjustment step, the fine adjustment step, and the coarse adjustment trigger threshold are set.

[0038] The fan regulation direction is set: when the differences of the three core water quality indexes are all positive values, the instruction to increase the air volume is sent to the fan.

[0039] When any of the differences of the three core water quality indexes is negative, the instruction to reduce the air volume is sent to the fan.

[0040] If any of the differences of the three core water quality indexes exceeds the coarse adjustment trigger threshold, the air volume is adjusted by the coarse adjustment step, and after the adjustment, the measured values of the three core water quality indexes in the non-interference water sample are re-collected based on the sampling period and the differences are calculated, and the coarse adjustment operation is repeated until the differences of the three core water quality indexes are all less than or equal to the coarse adjustment trigger threshold.

[0041] At this time, the air volume is fine-tuned by the fine adjustment step, and after the adjustment, the measured values of the three core water quality indexes in the non-interference water sample are re-collected and the differences are calculated, and if any of the differences of the three core water quality indexes reverses, the regulation operation is performed in the reverse direction by the fine adjustment step, until the differences of the three core water quality indexes are all less than or equal to the precision threshold, and the air volume adjustment is stopped.

[0042] Preferably, the specific process of integrating the aeration regulation direction and the corresponding stage correction step to generate the fan speed or frequency instruction and execute aeration is as follows:

[0043] The determined basic regulation direction and the correction step of the corresponding regulation stage are integrated.

[0044] The corresponding logic of the correction step in the adjustment direction is converted into the specific control parameters of the fan speed or frequency, generating executable fan speed / frequency adjustment instructions, according to which the actual aeration operation is performed.

[0045] Preferably, in the new aeration period, the optimal adaptive working condition template is screened, and the initial proportionality coefficient and the coarse and fine adjustment steps are extracted as the specific process of the initial parameters corresponding to the new working condition.

[0046] The full amount of core data of the current aeration adjustment is stored in the system database in a preset format, and a working condition comparison template for the current aeration adjustment is constructed.

[0047] The aeration adjustment working condition comparison template includes: core water quality characteristic vector and working condition adjustment parameter set.

[0048] In the new aeration period, the measured value of the core water quality index of the current undisturbed water sample is first obtained and normalized to construct the core water quality characteristic vector, and the cosine similarity is calculated with the core water quality characteristic vector of the database template to screen the high matching degree working condition template set.

[0049] From the high matching degree working condition template set that has been screened, the initial proportionality coefficient, coarse adjustment step threshold, and proportionality coefficient amplitude of all templates are extracted, and the adaptive scores of each evaluation parameter in each high matching degree working condition template are calculated, and the working condition adaptive score value corresponding to each high matching degree working condition template is calculated.

[0050] The high matching degree working condition template with the highest working condition adaptive score is selected and recorded as the optimal adaptive working condition template, and the initial proportionality coefficient and the coarse / fine adjustment step corresponding to the dynamic target value of the template are extracted as the initial corresponding parameters for calculating the water quality dynamic target value and determining the coarse and fine adjustment rules.

[0051] Preferably, the specific process of performing low-quality template purification operation on the aeration adjustment working condition comparison template in the database is as follows:

[0052] All aeration adjustment working condition comparison templates stored in the database are obtained, the core water quality characteristic vectors of each template are extracted, and the cosine similarity between the core water quality characteristic vectors of any two templates is calculated.

[0053] Based on the similarity result, a clustering algorithm is used to cluster and divide all templates to form several working condition template clusters, and a preset threshold of the number of templates in a cluster is set.

[0054] For each working condition template cluster, the total number of aeration adjustment working condition comparison templates contained is counted; if the number of templates in a cluster exceeds the preset threshold of the number of clusters, the template purification optimization process of the cluster is triggered, which is specifically as follows:

[0055] The water quality feedback results corresponding to the templates in the cluster are extracted, and a preset feedback score is extracted therefrom.

[0056] Ranking the templates in the cluster from low to high feedback score, removing the templates with the lowest feedback score after ranking and the number exceeding the preset removal threshold;

[0057] Reintegrating the remaining high-quality templates after removing low-quality templates, and integrating them into a new working condition template cluster to complete the purification iteration of the template cluster under this working condition.

[0058] Compared with the prior art, the beneficial effects of the present application are:

[0059] (1) The AI intelligent aeration method based on sewage treatment can accurately quantify the overall pollution state of the membrane assembly by deploying a self-cleaning rapid membrane separation device at the front end of the secondary sedimentation tank, calculating the membrane pollution characterization value of the membrane assembly, and avoiding pollution misjudgment caused by data deviation of a single detection point. In combination with the membrane pollution early warning evaluation value, online backwashing can be triggered in time to prevent excessive accumulation of membrane pollution from affecting separation efficiency and to continuously provide a stable equipment environment for non-interference water sample collection, thereby solving the problem of "uncertainty" of traditional sensors caused by interference of sludge flocs and suspended particles. At the same time, the pre-sampling of these pollution parameters allows the artificial intelligence to capture water quality and biochemical reaction trends before the end of the effluent, thereby avoiding the need to maintain a high aeration margin to avoid lag risks and greatly reducing energy consumption redundancy caused by "after-the-fact remediation".

[0060] (2) The AI intelligent aeration method based on sewage treatment sets an initial dynamic target value that is stricter than the national standard based on the national standard emission value, triggers target value optimization in combination with the situation of continuous compliance of water quality deviation, and gradually relaxes to approach the national standard to avoid aeration redundancy caused by "excessive conservatism" of traditional static targets. The two-way adjustment logic of "the core water quality index difference is positive to increase aeration, and any negative to decrease aeration" replaces the traditional single increase mode to eliminate invalid aeration from the adjustment direction. At the same time, relying on the synergistic strategy of coarse adjustment step to cope with load fluctuations and fine adjustment step to ensure steady-state accuracy, the problem of single step "oscillation or drift" is avoided, and the problem of perennial over-aeration is solved, thereby achieving dynamic matching of aeration demand and water quality compliance demand.

[0061] (3) The AI intelligent aeration method based on sewage treatment converts the total core data of single aeration into standardized working condition templates containing water quality feature vectors and adjustment parameter sets, selects the optimal template through cosine similarity matching and adaptation score in a new aeration cycle, extracts the initial parameters for direct reuse, greatly shortens the time-consuming of new working condition debugging, and reduces invalid energy consumption in the debugging phase. At the same time, low-quality templates are removed through clustering and scoring purification mechanism to realize continuous purification and optimization of the template library, provide high-quality data support for long-term stable operation of AI intelligent aeration, form a full-process closed loop of "adjustment-precipitation-reuse-optimization", and ensure that the aeration adjustment accuracy continues to improve with the running time. Attached Figure Description

[0062] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0063] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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 are within the scope of protection of the present invention.

[0064] Example 1;

[0065] Please see Figure 1 This invention provides an AI-powered intelligent aeration method for wastewater treatment, comprising:

[0066] Step 1: Deploy a rapid membrane separation device before the secondary sedimentation tank influent, analyze the membrane fouling characterization values ​​of the membrane module, construct a membrane fouling trend curve based on this, calculate the membrane fouling early warning assessment value, determine whether online backwashing is triggered, collect undisturbed water samples, and obtain the core water quality indicators of the undisturbed water samples. The specific process is as follows:

[0067] A self-cleaning small-scale rapid membrane separation device is deployed before the secondary sedimentation tank influent. The device has a built-in online backwashing unit, and several detection points are evenly distributed on the surface of the membrane module corresponding to the membrane flux monitoring sensor along the sewage flow direction. Each detection point is equipped with a transmembrane differential pressure sensor.

[0068] Set the data acquisition period. For each detection point, use a transmembrane differential pressure sensor to obtain the transmembrane differential pressure value on the membrane surface at the acquisition time, and record it as the membrane fouling characterization value S at the detection point.

[0069] The detection points are sorted according to the direction of wastewater flow, and a basic weight is preset for each detection point according to the order of the detection points along the water flow direction. The detection points closer to the water inlet side come into contact with wastewater first, the pollutants attach earlier, and the degree of pollution is usually more significant. Their data has higher reference value for judging the overall pollution status of the membrane module. The earlier the detection point is sorted, the greater the basic weight.

[0070] The basic weights of each detection point are normalized to obtain the pollution impact weight Wi of each detection point, where the sum of the pollution impact weights of all detection points is 1, i is the label of the detection point, i=1,2,...,n; n is the total number of detection points.

[0071] Obtain the membrane fouling characterization value Si and the corresponding fouling influence weight Wi at each detection point at the acquisition time, and use the formula: , obtain the membrane pollution characterization value MW of the membrane module of the rapid membrane separation equipment;

[0072] With time as the horizontal coordinate and the membrane pollution characterization value of the membrane module as the vertical coordinate, a rectangular coordinate system is constructed, the membrane pollution characterization value of the membrane module corresponding to each collection time is obtained with the last self-cleaning of the rapid membrane separation equipment as the initial time, and corresponding labeling is performed in the rectangular coordinate system, thereby obtaining a plurality of membrane pollution data points. According to the time sequence, the adjacent membrane pollution data points are sequentially connected through a smooth curve, thereby obtaining a membrane pollution trend change curve graph;

[0073] The cumulative pollution increment is obtained by subtracting the membrane pollution characterization value corresponding to the initial time from the membrane pollution characterization value corresponding to the current collection time. The unit time pollution increase rate AV is obtained by dividing the cumulative pollution increment by the total duration of the continuous collection period.

[0074] The maximum increment of the membrane pollution characterization value of the membrane module per unit time lasting to the current collection time is obtained, thereby obtaining the pollution maximum mutation amplitude AF.

[0075] After normalization and dimensionless processing of the membrane pollution characterization value MW of the membrane module corresponding to the current collection time, the unit time pollution increase rate AV, and the pollution maximum mutation amplitude AF,

[0076] Using the formula: , the membrane pollution early warning evaluation value YP is obtained, wherein b1, b2, and b3 are preset weight coefficients.

[0077] A membrane pollution early warning evaluation threshold is preset. If the membrane pollution early warning evaluation value is greater than or equal to the corresponding preset threshold, a pollution early warning is triggered, and an online backwashing unit is started for low-pressure backwashing. The membrane module continuously separates suspended particulate matter in sewage.

[0078] The non-interference clear water sample (interference impurities such as suspended particulate matter, colloid, and activated sludge floc in sewage are removed, and only clear water samples of dissolved pollutants and water basic characteristics are retained) is collected through the clear water collection channel of the self-cleaning small rapid membrane separation equipment at a preset sampling period. A rapid water quality measuring instrument and a multi-parameter sensor array are integrated at the outlet end of the clear water collection channel, and the core water quality indicators of the non-interference water sample are collected based on the sensor array.

[0079] The core water quality indicators include chemical oxygen demand, ammonia nitrogen, and total nitrogen.

[0080] Chemical oxygen demand (COD): It represents the total amount of organic matter in sewage that can be oxidized by chemical oxidants, and is a core indicator for measuring the degree of organic pollution of water bodies.

[0081] Ammonia nitrogen (NH3-N): It represents the degradation intermediate product of nitrogen-containing organic pollutants in sewage, and its removal effect is directly related to the nitrification efficiency of biochemical reactions.

[0082] Total nitrogen (TN): represents the total content of various nitrogen-containing compounds in wastewater, is the key indicator for controlling water eutrophication;

[0083] All collected data are subjected to noise filtering and data normalization preprocessing.

[0084] It should be noted that the core value of the rapid membrane separation device is to solve the problem of feedback lag in traditional wastewater treatment. In traditional process, the effluent of biochemical tank needs to pass through multiple links of sludge and water separation such as secondary sedimentation tank, sand filter or magnetic coagulation, high-density sedimentation tank, etc. The whole process takes several hours. When the water quality is detected and the air volume adjustment is determined to be reasonable, the best adjustment opportunity has been missed. However, the device can complete the equivalent separation effect in a few minutes, quickly obtain the interference-free water sample free of suspended particles, colloids and other interference impurities, so that the AI can verify whether the air volume adjustment is suitable for the water quality demand in real time, solve the problem of "after-the-fact remedy" type aeration from the source, and ensure the timeliness of the adjustment response.

[0085] The self-cleaning micro-membrane assembly physically intercepts suspended solids, colloids and activated sludge flocs at the front end of the secondary sedimentation tank, allowing only dissolved pollutants to penetrate, so that the core water quality indicators are free from solid interference, and the measurement stability is significantly enhanced, providing low-noise and high-consistency input samples for subsequent processes and laying the foundation for accurate control.

[0086] The transmembrane pressure difference sensing array arranged at multiple points along the water flow direction generates a membrane pollution representation value through a normalized weighting strategy, and further constructs an early warning evaluation index by combining the pollution growth rate and mutation amplitude. When the index exceeds the preset threshold, the low-pressure ultrasonic online backwashing is triggered immediately to realize the online recovery of membrane flux and avoid the signal interruption and manual input caused by traditional offline cleaning, thereby ensuring continuous monitoring.

[0087] The interference-free water sample collection point is located at the front section of the secondary sedimentation tank, so that the water quality and biochemical state information are captured before the effluent is formed, and the trend prediction and aeration decision can be completed before the final discharge of pollutants, thereby significantly reducing the feedback lag and improving the response speed of the system to load fluctuations.

[0088] Step 2: Obtain the measured values of the core water quality indicators of the interference-free water sample and the national standard values, and calculate the dynamic target values of each water quality indicator. Take the dynamic target value as the initial control target, adjust the air volume through the fan adjustment mechanism, monitor the measured values of the water quality indicators, calculate the difference values of the water quality indicators, and determine whether to trigger the optimization and iteration of the dynamic target value. Determine the adjustment direction of the fan and the coarse adjustment and fine adjustment rules until the deviation of the water quality indicators meets the accuracy requirements. The specific process is as follows:

[0089] Obtain the measured values of the core water quality indicators of the interference-free water sample, specifically: chemical oxygen demand, ammonia nitrogen and total nitrogen, respectively denoted as chemical oxygen demand measured value , ammonia nitrogen measured value , measured value of total nitrogen ;

[0090] Obtain the national emission standard value of each core water quality index, respectively denoted as chemical oxygen demand national standard value , ammonia nitrogen national standard value , total nitrogen national standard value ;

[0091] preset initial proportion coefficient , by multiplying the national emission standard value of each core water quality index by the preset initial proportion coefficient, the dynamic target value of each core water quality index is obtained, respectively denoted as chemical oxygen demand dynamic target value , ammonia nitrogen dynamic target value , total nitrogen dynamic target value , the corresponding formula is:

[0092] ;

[0093] Among them, each dynamic target value is more stringent than the corresponding national standard value;

[0094] Take the dynamic target value as the initial control target, continuously aerate and adjust the sewage to be treated in the biochemical tank through the fan air volume adjustment mechanism, and monitor the measured values of chemical oxygen demand, ammonia nitrogen and total nitrogen in the interference-free water sample collected by the rapid membrane separation equipment in each preset sampling period in real time;

[0095] For each sampling period, by subtracting the corresponding dynamic target value from the measured value of each core water quality index, the difference value of each core water quality index is obtained, respectively denoted as chemical oxygen demand difference value , ammonia nitrogen difference value , total nitrogen difference value , the corresponding formula is:

[0096] ;

[0097] Among them, when the difference value is positive, the corresponding measured value is greater than the dynamic target value, and the corresponding water quality index is worse than the target; when the difference value is negative, the corresponding measured value is less than the dynamic target value, and the water quality is better than the target;

[0098] preset precision threshold , proportion coefficient increment , proportion coefficient final value , and the number of consecutive compliance periods k;

[0099] If the current sampling period satisfies: (that is, the deviation of the measured value of each core water quality index and the dynamic target value is controlled within the precision threshold), and on this basis, the continuous k sampling periods remain in this state, the dynamic target value optimization is triggered;

[0100] The specific process of dynamic target value optimization is as follows:

[0101] By multiplying the current proportion coefficient with the proportion coefficient increment , the updated proportion coefficient is obtained:

[0102] wherein m is the index of the sampling period;

[0103] The dynamic target value of each core water quality index is updated based on the dynamic target value calculation formula, so that the target value is gradually relaxed and approaches the corresponding national standard value;

[0104] The proportion coefficient is repeatedly updated by the proportion coefficient increment, the new dynamic target value is calculated based on the updated proportion coefficient, and it is monitored in real time whether the water quality deviation in each sampling period meets the following conditions: to trigger the next optimization cycle until the proportion coefficient reaches the preset proportion coefficient final value Xend%, the dynamic target value iteration is stopped, and the current dynamic target value is kept stable;

[0105] The specific process of the fan air volume regulation mechanism is as follows:

[0106] The preset coarse adjustment step , fine adjustment step and coarse adjustment trigger threshold δ are set;

[0107] The fan regulation direction is set as:

[0108] When is met at the same time, the fan is sent an instruction to increase the air volume (i.e., all water quality indicators are not up to standard, and the fan air volume needs to be increased to increase aeration and degrade pollutants);

[0109] When any of them is met, the fan is sent an instruction to reduce the air volume (i.e., at least one core indicator is less than the dynamic target value, and the water quality is better than the target. At this time, because as long as one indicator meets the standard and has redundancy, there is no need to maintain high air volume aeration to avoid waste of aeration);

[0110] If any of them is met, the air volume is adjusted by the coarse adjustment step, and the water quality indicator measured value is re-collected based on the sampling period and the difference is calculated. Repeat the coarse adjustment operation until is met at the same time;

[0111] When At the same time, switch to fine adjustment step fine adjustment air volume, after adjustment, based on the sampling period to collect water quality index measured value and calculate the difference, if the fine adjustment of any core water quality index difference direction is reversed (positive to negative or negative to positive), the reverse direction is executed according to the fine adjustment step (for example: reason Execute fine adjustment of increasing air volume, ΔC changes from positive to negative after fine adjustment, at this time, execute fine adjustment of reducing air volume in reverse direction, and the adjustment amplitude is still the fine adjustment step), until At the same time, stop air volume adjustment.

[0112] It should be noted that the national standard emission value is taken as the upper limit, the dynamic target value stricter than the upper limit is formed according to the preset initial proportion coefficient, and the deviation is calculated based on the dynamic target value; when the deviation is continuously better than the precision threshold and maintained for a set period, the dynamic target value is automatically relaxed by the preset increment until it approaches the national standard value. In this way, the system can gradually tap the emission reduction potential under the premise of ensuring the safety redundancy of effluent, and realize the self-evolution control from "overly conservative" to "lean margin";

[0113] When all core water quality index deviations are positive, it is determined that the aeration is insufficient, and the instruction to increase air volume is executed; when any index deviation is negative, that is, there is a standard redundancy, the instruction to reduce air volume is executed, thereby eliminating "one-size-fits-all" high aeration operation. This strategy makes the air volume follow the real demand and float in both directions, for the first time, "reducing aeration" is included in the normal control logic, which significantly reduces invalid energy consumption;

[0114] Set the coarse adjustment trigger threshold, and any index deviation exceeding the threshold will enable the larger step to quickly flatten; when all indexes return to within the threshold, automatically switch to the smaller step for fine convergence until the precision threshold is met. The coarse level guarantees dynamic response speed, and the fine level guarantees steady-state control accuracy, avoiding the inherent defects of "too fast oscillation" or "too slow drift" of single-stage regulation;

[0115] The step-by-step relaxation of the dynamic target value creates a safe space for air volume reduction; the water quality improvement caused by air volume adjustment provides data support for the next round of target value relaxation; the two are iterated alternately to form a positive cycle of "target relaxation-aeration reduction-water quality continuous standard", which realizes the step-by-step reduction of aeration energy consumption under the premise that the effluent water quality is always better than the national standard.

[0116] Step three: integrate the aeration adjustment direction and the corresponding stage correction step to generate fan speed or frequency instruction, and execute aeration operation; store all core data of aeration adjustment and build aeration adjustment working condition comparison template; in the next round of aeration adjustment period, select the optimal adaptive working condition template, and based on the corresponding dynamic target value initial proportion coefficient and coarse / fine adjustment step, as the initial reference parameters of the next round of aeration adjustment; perform low-quality template purification operation on the aeration adjustment working condition comparison template in the database, and the specific process is:

[0117] Integrating the base adjustment direction (increase or decrease air volume) determined in step two and the correction step length of the corresponding adjustment stage (i.e., the coarse adjustment step length corrected by the biochemical reaction state in the coarse adjustment stage , and the fine adjustment step length corrected by the secondary energy consumption in the fine adjustment stage );

[0118] According to the corresponding logic of the adjustment direction + the correction step length, the specific control parameters of the fan speed / frequency are generated to form executable fan speed / frequency adjustment instructions, which are synchronously pushed to the local PLC controller by artificial intelligence (Ai), and then the linkage control between the PLC controller and the fan execution unit is realized to land the optimized adjustment logic to the actual aeration operation, so as to realize the intelligent control of the aeration process by artificial intelligence (Ai);

[0119] The full amount of core data of the current aeration adjustment is stored in the system database according to the preset format, including: the measured values of the core water quality indicators (chemical oxygen demand, ammonia nitrogen, and total nitrogen) of the undisturbed water sample collected in step one; the dynamic target values of each water quality indicator, the water quality indicator difference, and the coarse / fine adjustment step length parameters in step two; the proportion coefficient of dynamic target value optimization and the number of consecutive standard periods; and the final fan control instruction of this round and the water quality feedback result after the adjustment is completed, and based on the above stored data, an aeration adjustment working condition comparison template is constructed;

[0120] Each template contains two core modules: a core water quality feature vector and a working condition adjustment parameter set, wherein the core water quality feature vector is formed by the normalized processing of the measured values of chemical oxygen demand, ammonia nitrogen, and total nitrogen in step one, and the working condition adjustment parameter set includes: the initial proportion coefficient, the dynamic target values of each water quality indicator, the water quality indicator difference, the coarse adjustment step length, the fine adjustment step length, the coarse adjustment trigger threshold, the precision threshold, the proportion coefficient increment, the number of consecutive standard periods, and the final fan adjustment direction (increase air volume or decrease air volume) and the corresponding stage air volume adjustment execution parameters determined in step two;

[0121] When entering a new aeration adjustment period, before performing step two to calculate the dynamic target values of each water quality indicator and determine the coarse adjustment and fine adjustment rules, the historical big data calling process of the database is automatically triggered, which is as follows:

[0122] The measured values of the core water quality indicators of the undisturbed water sample corresponding to the new aeration adjustment period are obtained, and the current core water quality feature vector is constructed after normalization processing;

[0123] The current core water quality feature vector is substituted into the system database, and the cosine similarity calculation is performed with the corresponding core water quality feature vectors in all aeration adjustment working condition comparison templates;

[0124] All aeration adjustment working condition comparison templates with a cosine similarity greater than a preset threshold are selected and integrated into a high-matching-degree working condition template set;

[0125] From the high matching condition template set that has been screened, the initial proportion coefficient, the coarse adjustment step threshold, and the proportion coefficient increment of all templates are extracted, and the abnormal values of the parameters caused by equipment failure and abnormal conditions are removed;

[0126] The effective historical data of each type of parameter is statistically analyzed, and the 25th percentile to the 75th percentile of the data is taken as the optimal fitting interval of the evaluation parameter, and the interval median QZ and interval half length QB of each type of evaluation parameter optimal fitting interval are calculated simultaneously;

[0127] For each evaluation parameter in each high matching condition template, the formula is used: , to obtain the evaluation parameter fitting score CP, wherein CS is the actual value of the corresponding evaluation parameter in the current high matching condition template;

[0128] The formula is used: , to obtain the condition fitting score GP, wherein f is the label of the initial proportion coefficient, the coarse adjustment step threshold, and the proportion coefficient increment, is the preset weight coefficient corresponding to the fth evaluation parameter;

[0129] From the high matching condition template set, the high matching condition template with the highest condition fitting score is selected, and is recorded as the optimal fitting condition template. The initial proportion coefficient and the coarse adjustment / precision adjustment step corresponding to the dynamic target value of the template are extracted as the initial corresponding parameters for calculating the water quality dynamic target value and determining the coarse adjustment and precision adjustment rules in step two, realizing the adaptation and reuse of historical high-quality adjustment experience for new conditions;

[0130] All aeration adjustment condition comparison templates stored in the system database are obtained, the core water quality feature vectors of each template are extracted, and the cosine similarity between the core water quality feature vectors of any two templates is calculated. Based on the cosine similarity result, a clustering algorithm is used to cluster and divide all templates to form several condition template clusters, and a preset threshold for the number of templates in a cluster is set, wherein each condition template cluster corresponds to an aeration adjustment condition with similar water quality characteristics;

[0131] For each generated condition template cluster, the total number of aeration adjustment condition comparison templates contained in it is counted. If the number of templates corresponding to a condition template cluster is greater than the preset threshold for the number of clusters, the template purification optimization process of the cluster is triggered, which is specifically:

[0132] The water quality feedback results corresponding to each aeration adjustment condition comparison template in the current condition template cluster are extracted, and the corresponding preset feedback score (the preset feedback score can be calculated by weighting the quantitative indicators of the water quality standard reaching dimension, energy consumption control dimension, and adjustment stability dimension and the preset weights by the staff) is extracted from the water quality feedback results;

[0133] The templates in the cluster are ranked from low to high according to the feedback score, and the aeration adjustment working condition comparison templates with the lowest feedback score after ranking and the number exceeding the preset removal threshold are removed;

[0134] The high-quality templates remaining after removing the low-quality templates are re-integrated into a new working condition template cluster, and the purification iteration of the template cluster under this type of working condition is completed accordingly, improving the overall precision and adaptation value of the templates in the cluster.

[0135] It should be noted that this process converts the adjustment direction and correction step determined in step two into executable instructions for fan speed / frequency, and through AI linkage with PLC controllers and fan execution units, it eliminates the lag and subjectivity of traditional manual adjustment, allowing the aeration adjustment logic to be accurately implemented, ensuring real-time matching of aeration intensity and water quality requirements in the biochemical tank. Relying on the rapid computing power of AI, the response efficiency of aeration adjustment is greatly improved, avoiding water quality fluctuations caused by delayed instruction transmission or execution;

[0136] The aeration adjustment working condition comparison templates constructed based on full-quantity core data achieve the standardization and data-based retention of single aeration adjustment experience. The cosine similarity matching and parameter adaptation score screening process of the new cycle can quickly locate the historical high-quality working condition that best adapts to the current water quality characteristics. The initial proportion coefficient and coarse / fine adjustment step extracted can provide accurate initial parameters for new working condition adjustment, eliminating the need for zero-start debugging, which not only shortens the time required for aeration adjustment to meet the standards, but also reduces the energy consumption of invalid aeration during the debugging phase, while avoiding the risk of water quality exceeding standards caused by blind parameter setting in new working conditions;

[0137] For the purification iteration of the database template, the clustering algorithm is used to classify homogeneous working condition templates, and the low-quality templates are removed based on multi-dimensional feedback scores, which can continuously purify the template library, avoid interference from invalid and incorrect working condition data in the new cycle, and improve the accuracy and adaptability of the template library as the running time increases. This provides reliable data support for the long-term stable and efficient operation of AI intelligent aeration, and also provides high-quality reference for the working condition migration of similar subsequent sewage treatment scenarios;

[0138] From aeration instruction execution, data storage and template modeling to new working condition experience reuse and template purification, step three links the "execution - retention - reuse - optimization" full link of aeration adjustment, forming a technical synergy with the accurate monitoring of step one and the dynamic target value iterative adjustment of step two, and constructing a complete closed loop from front-end data acquisition to end working condition optimization. This not only ensures the stable and standard water quality, but also achieves continuous reduction of aeration energy consumption, achieving the dual goals of "water quality standard" and "energy saving and consumption reduction" in sewage treatment.

[0139] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications and adaptions can be effected without departing from the spirit and scope of the present application, which is not limited to the exact construction and arrangement described. It is intended, therefore, to cover all modifications and adaptions that fall within the scope of the claims and their equivalents.

Claims

1. AI intelligent aeration method based on sewage treatment, characterized in that, Comprising the following steps: Step one: Deploy a rapid membrane separation device at the front of the secondary sedimentation tank before water intake, analyze the membrane pollution characterization value of the membrane module, construct a membrane pollution trend change curve based on the value, calculate the membrane pollution early warning evaluation value, judge whether to trigger online backwashing, collect non-interference water samples, and obtain the core water quality indicators of non-interference water samples; Step two: Obtain the measured value of the core water quality indicators of non-interference water samples, and calculate the dynamic target value of each water quality indicator; Take the dynamic target value as the initial control target, adjust the aeration amount through the fan air volume adjustment mechanism and monitor the measured value of the water quality indicators, calculate the difference value of the water quality indicators, judge whether to trigger the dynamic target value optimization and iteration, and determine the fan adjustment direction and coarse adjustment and fine adjustment rules until the water quality indicator deviation meets the accuracy requirement; The specific process of the fan air volume adjustment mechanism is as follows: Pre-set coarse adjustment step, fine adjustment step and coarse adjustment trigger threshold; Set the fan adjustment direction: when the difference value of the three core water quality indicators is positive, send the instruction to increase the air volume to the fan; When any of the difference values of the three core water quality indicators is negative, send the instruction to reduce the air volume to the fan; If any of the difference values of the three core water quality indicators exceeds the coarse adjustment trigger threshold, adjust the air volume by the coarse adjustment step, and then collect the measured values of the three core water quality indicators in the non-interference water samples based on the sampling period and calculate the difference values, and repeat the coarse adjustment operation until the difference values of the three core water quality indicators are all less than or equal to the coarse adjustment trigger threshold; At this time, switch to fine adjustment step to fine-tune the air volume, and then collect the measured values of the three core water quality indicators in the non-interference water samples and calculate the difference values, if any of the difference values of the three core water quality indicators reverses, execute the adjustment operation in the reverse direction by the fine adjustment step, until the difference values of the three core water quality indicators are all less than or equal to the accuracy threshold, and stop adjusting the air volume; Step three: Integrate the aeration adjustment direction and the corresponding stage correction step to generate the fan speed or frequency instruction and execute the aeration; Store the full amount of core data of aeration adjustment, construct the aeration adjustment working condition comparison template; in the new aeration period, select the optimal adaptive working condition template, extract its initial proportion coefficient and coarse adjustment and fine adjustment step as the initial parameters corresponding to the new working condition; perform low-quality template purification operation on the templates in the database; The specific process of performing low-quality template purification operation on the aeration adjustment working condition comparison templates in the database is as follows: Obtain all the stored aeration adjustment working condition comparison templates in the database, extract the core water quality feature vectors of each template, and calculate the cosine similarity between the core water quality feature vectors of any two templates; Based on the similarity results, use clustering algorithm to cluster all templates to form several working condition template clusters, and pre-set the threshold of the number of templates in each cluster; For each working condition template cluster, count the total number of aeration adjustment working condition comparison templates it contains; if the number of templates in a cluster exceeds the pre-set threshold of the number of clusters, trigger the template purification optimization process of the cluster, which is as follows: Extract the water quality feedback results corresponding to each template in the cluster, and extract the pre-set feedback score from them; Sort the templates in the cluster from low to high according to the feedback score, and remove the templates with the lowest feedback score after sorting and whose number exceeds the pre-set removal threshold; Re-integrate the remaining high-quality templates after removing the low-quality templates, and integrate them into a new working condition template cluster to complete the purification iteration of the template cluster under this type of working condition.

2. The AI intelligent aeration method based on sewage treatment according to claim 1, characterized in that: The specific process of analyzing the membrane pollution characterization value of the membrane assembly is as follows: A self-cleaning small rapid membrane separation device is arranged in the front of the secondary sedimentation tank before the influent is introduced, the device is provided with an online backwashing unit, and a plurality of detection points are uniformly arranged on the surface of the membrane assembly corresponding to the rapid membrane separation device along the sewage flow direction; A collection period is set, the transmembrane pressure difference value of the membrane surface of each detection point is obtained, and the detection point membrane pollution characterization value is recorded; The detection points are sorted according to the sewage flow direction and a basic weight is preset, and then the basic weight is normalized to obtain the pollution influence weight of each detection point; The membrane pollution characterization value of the membrane assembly corresponding to the rapid membrane separation device is obtained by combining the membrane pollution characterization value of each detection point and the corresponding pollution influence weight and weighted calculation.

3. The AI intelligent aeration method based on sewage treatment according to claim 2, characterized in that: The specific process of calculating the membrane pollution early warning evaluation value and judging whether the online backwashing is triggered is as follows: A rectangular coordinate system is constructed with time as the horizontal coordinate and the membrane pollution characterization value of the membrane assembly as the vertical coordinate, the membrane pollution characterization value of the membrane assembly at each collection time is marked with the last self-cleaning of the rapid membrane separation device as the initial time, the pollution data points are obtained, and the membrane pollution trend change curve is connected in time sequence with a smooth curve to form a curve diagram; The unit time pollution increase rate is obtained by subtracting the membrane pollution characterization value of the membrane assembly at the initial time from the membrane pollution characterization value of the membrane assembly at the current collection time and then dividing by the total duration of the continuous collection period; The maximum increment per unit time of the membrane pollution characterization value of the membrane assembly from the initial time to the current time is taken as the maximum mutation amplitude of the pollution; The membrane pollution early warning evaluation value is obtained by weighted analysis of the membrane pollution characterization value of the membrane assembly at the current collection time, the unit time pollution increase rate and the maximum mutation amplitude of the pollution; If the membrane pollution early warning evaluation value is greater than or equal to the corresponding preset threshold value, the pollution early warning is triggered, and the online backwashing unit is started for low-pressure backwashing.

4. The AI intelligent aeration method based on sewage treatment according to claim 3, characterized in that: The specific process of collecting non-interference water samples and obtaining the core water quality indicators and biochemical reaction state indicators of the non-interference water samples is as follows: Non-interference clean water samples are collected through the clean water collection channel of the self-cleaning small rapid membrane separation device at a preset sampling period, and the core water quality indicators of the non-interference water samples are collected; The water quality indicators include chemical oxygen demand, ammonia nitrogen and total nitrogen.

5. The AI intelligent aeration method based on sewage treatment according to claim 4, characterized in that: The specific process of monitoring the measured values of the water quality indicators and calculating the difference values of the water quality indicators is as follows: The measured values of the core water quality indicators corresponding to the non-interference water samples are obtained; The national discharge standard values of each core water quality indicator are obtained; An initial proportion coefficient is preset, and the dynamic target value of each core water quality indicator is obtained by multiplying the national discharge standard value of each core water quality indicator by the preset initial proportion coefficient; The dynamic target value is taken as the initial control target, the sewage to be treated in the biochemical tank is continuously aerated and adjusted through the fan air volume adjustment mechanism, and the measured values of the water quality indicators in the non-interference water samples collected by the rapid membrane separation device in each preset sampling period are monitored in real time; For each sampling period, the difference value of each core water quality indicator is obtained by subtracting the corresponding dynamic target value from the measured value of each core water quality indicator.

6. The AI intelligent aeration method based on sewage treatment according to claim 5, characterized in that: The specific process of judging whether the dynamic target value optimization and iteration are triggered is as follows: An accuracy threshold, a proportion coefficient increment, a proportion coefficient final value and a continuous compliance period number are preset. If the absolute values of the differences of the three core water quality indicators in the current sampling period are all less than or equal to the preset precision threshold, and this state is maintained for a continuous period, the dynamic target value optimization is triggered. The specific process of dynamic target value optimization is as follows: The updated proportional coefficient is obtained by adding the proportional coefficient increment to the current proportional coefficient, and the dynamic target value is updated based on the new coefficient. The proportional coefficient is updated by the proportional coefficient increment, the new dynamic target value is calculated, and it is monitored whether the water quality deviation in each collection period meets the condition that the absolute values of the differences of the three water quality indicators in the continuous period are all less than or equal to the preset precision threshold, so as to trigger the next optimization cycle, until the proportional coefficient reaches the preset proportional coefficient final value, the dynamic target value iteration is stopped, and the current dynamic target value is kept stable.

7. The AI intelligent aeration method based on sewage treatment according to claim 6, characterized in that: The specific process of integrating the aeration adjustment direction and the corresponding stage correction step to generate the fan speed or frequency command and executing aeration is as follows: Integrate the determined basic adjustment direction and the correction step of the corresponding adjustment stage. According to the specific control parameters of the fan speed or frequency corresponding to the adjustment direction + correction step, the executable fan speed / frequency adjustment command is generated, and the actual aeration operation is executed accordingly.

8. The AI intelligent aeration method based on sewage treatment according to claim 7, characterized in that: In the new aeration period, the optimal adaptive working condition template is selected, and the initial proportional coefficient, coarse adjustment step and fine adjustment step are extracted as the initial parameters corresponding to the new working condition. The specific process is as follows: Store the full amount of core data of this round of aeration adjustment in the system database according to the preset format, and build a comparison template of this round of aeration adjustment working condition. The aeration adjustment working condition comparison template includes: core water quality feature vector and working condition adjustment parameter set. In the new aeration period, the core water quality index measured value of the current undisturbed water sample is first obtained and normalized to build the core water quality feature vector, and the cosine similarity is calculated with the core water quality feature vector of the database template to form a high matching degree working condition template set. From the selected high matching degree working condition template set, the initial proportional coefficient, coarse adjustment step threshold and proportional coefficient increment of all templates are extracted, and the adaptive scores of each evaluation parameter in each high matching degree working condition template are calculated, and the working condition adaptation score value corresponding to each high matching degree working condition template is calculated. The high matching degree working condition template with the highest adaptation score is selected and recorded as the optimal adaptive working condition template, and the initial proportional coefficient and coarse / fine adjustment step corresponding to the template are extracted as the initial corresponding parameters for calculating the water quality dynamic target value and determining the coarse / fine adjustment rule.

Images