Sewage treatment plant carbon management and control method and system based on artificial intelligence

Abstract

The invention discloses a sewage treatment plant carbon management and control method and system based on artificial intelligence. The method comprises the steps that a carbon emission prediction model is constructed according to collected real-time monitoring data, historical data and a mechanism model; generating a parameter optimization trigger instruction based on the processing load data and the water condition change rate output by the carbon emission prediction model, and further generating a process unit operation parameter configuration scheme based on the water condition; obtaining a first carbon emission predicted value corresponding to the process unit operation parameter configuration scheme, and establishing a dynamic parameter regulation and control strategy library; and matching the real-time water condition fluctuation data with the dynamic parameter regulation and control strategy library and the mechanism model, determining a real-time parameter regulation and control strategy, and executing an equipment parameter adjustment instruction according to the real-time parameter regulation and control strategy. The problem that a current sewage treatment plant cannot predict the carbon emission conditions under different water conditions in real time is solved, and the effects of predicting the carbon emission conditions under different water conditions in real time and reducing energy loss are achieved.

Description

TECHNICAL FIELD

[0001] The application relates to the sewage treatment technical field, in particular to a sewage treatment plant carbon control method and system based on artificial intelligence. BACKGROUND

[0002] As an important part of urban infrastructure, sewage treatment is not only an important source of carbon emissions in ensuring water environment safety, but also directly affects the realization process of the national carbon peak and carbon neutralization target.

[0003] With the increasingly stringent environmental protection requirements and the increasing pressure of carbon emission reduction, carbon emission control of sewage treatment plants has become a key issue in environmental governance. At present, the carbon emission control of sewage treatment plants mainly relies on artificial experience and fixed parameter setting, which is difficult to respond to the dynamic changes of water treatment capacity in real time. Moreover, due to the use of static operation mode, relevant technical personnel cannot flexibly adjust equipment parameters according to actual treatment load, which not only causes energy waste, but also leads to a continuous high carbon emission. In order to solve the above problems, relevant technical personnel usually use post hoc statistical analysis method to collect and arrange historical operation data, calculate key indicators such as energy consumption carbon emission, and then evaluate the carbon emission reduction effect based on the statistical results.

[0004] However, this passive management mode of "first emission, then statistics" has obvious hysteresis, and sewage treatment plants are difficult to develop corresponding parameter adjustment strategies in advance and can only passively respond to changes in water quantity and quality. SUMMARY

[0005] The embodiment of the application provides a sewage treatment plant carbon control method based on artificial intelligence, which solves the problem that the sewage treatment plant cannot predict the carbon emission under different water conditions in real time, realizes real-time prediction of carbon emission under different water conditions, reduces unnecessary energy consumption, and reduces carbon emission during sewage treatment.

[0006] The embodiment of the application provides a sewage treatment plant carbon control method based on artificial intelligence, which comprises: According to the collected real-time monitoring data, historical data and mechanism model, a carbon emission prediction model is constructed, wherein the real-time monitoring data comprises water condition fluctuation data and process unit running state data; the water condition fluctuation data comprises water quantity fluctuation data and water quality fluctuation data; the historical data comprises historical water condition fluctuation sequence, historical process unit running parameter and historical carbon emission; A parameter optimization trigger instruction is generated based on the processing load data and water condition change rate output by the carbon emission prediction model, and a process unit running parameter configuration scheme based on water condition is generated according to the parameter optimization trigger instruction; obtaining a first carbon emission prediction value corresponding to the process unit operation parameter configuration scheme, and establishing a dynamic parameter control strategy library; matching the real-time water condition fluctuation data with the dynamic parameter control strategy library and the mechanism model to determine a real-time parameter control strategy, and executing a device parameter adjustment instruction according to the real-time parameter control strategy.

[0007] Optionally, the step of generating a parameter optimization trigger instruction based on the processing load data and the water condition change rate output by the carbon emission prediction model, and generating a process unit operation parameter configuration scheme based on a water condition comprises: if the carbon emission value in the processing load data exceeds a preset carbon emission threshold value, and the water condition change rate exceeds a preset fluctuation amplitude, the parameter optimization trigger instruction is generated; starting a genetic algorithm program according to the parameter optimization trigger instruction, obtaining a running parameter boundary value of the process unit, and initializing a population individual matrix of the genetic algorithm according to the running parameter boundary value, wherein the population individual matrix comprises a device parameter combination of aeration quantity, reflux ratio and sludge concentration; evaluating a second carbon emission prediction value and a processing efficiency index of the device parameter combination based on the water condition, and generating a fitness score; if the fitness score does not meet an expected condition, performing a crossover and mutation operation on the device parameter combination, re-evaluating the device parameter combination after the crossover and mutation, and updating the fitness score; when the fitness score meets the expected condition, outputting the device parameter combination corresponding to the expected condition, and generating the process unit operation parameter configuration scheme.

[0008] Optionally, the step of constructing a carbon emission prediction model according to the collected real-time monitoring data, historical data and mechanism model comprises: taking the historical water condition fluctuation sequence as an input feature vector and the historical carbon emission sequence as a target output value to generate a training data set, wherein each training sample contains water condition data and a corresponding single carbon emission value of a preset time length; calculating a loss function value between a predicted carbon emission value and an actual carbon emission value based on the training data set by a back propagation algorithm; adjusting a network parameter configuration according to the loss function value, evaluating the prediction accuracy of the network parameter configuration, and determining a first weight parameter; receiving the real-time monitoring data according to the first weight parameter, constructing a prediction input vector, and generating the carbon emission prediction model in combination with the mechanism model.

[0009] Optionally, the step of obtaining the first carbon emission prediction value corresponding to the process unit operation parameter configuration scheme, and establishing a dynamic parameter regulation strategy library comprises: If the difference between the first carbon emission prediction value and the preset carbon emission threshold value is less than the critical range, the carbon emission threshold correlation degree of the process unit operation parameter configuration scheme is calculated; If the carbon emission threshold correlation degree exceeds the preset correlation reference value, an emergency regulation index is assigned, and a hierarchical index structure strategy classification result is generated; The storage structure of the dynamic parameter regulation strategy library is established through the strategy classification result, and a strategy retrieval mechanism is constructed according to the emergency regulation index, and the calling weight of the process unit operation parameter configuration scheme regulation scheme is determined; According to the calling weight, a hierarchical management dynamic parameter regulation strategy library is generated.

[0010] Optionally, the step of matching the real-time water quality fluctuation data with the dynamic parameter regulation strategy library and the mechanism model, determining a real-time parameter regulation strategy, and executing a device parameter adjustment instruction according to the real-time parameter regulation strategy comprises: Obtain the fluctuation feature vector of the water quality fluctuation data, and perform similarity matching of the fluctuation feature vector with the dynamic parameter regulation strategy library and the mechanism model, and extract a parameter regulation strategy record according to the matching result; Perform carbon emission prediction calculation on the device operation parameters in the parameter regulation strategy record, and if the predicted carbon emission value is less than a preset safety threshold, extract a device parameter adjustment instruction from the parameter regulation strategy record to determine the real-time parameter regulation strategy; According to the parameter type identifier in the real-time parameter regulation strategy, determine the target device control interface address, and send the device parameter adjustment instruction to the target device.

[0011] Optionally, after the step of matching the real-time water quality fluctuation data with the dynamic parameter regulation strategy library and the mechanism model, determining a real-time parameter regulation strategy, and executing a device parameter adjustment instruction according to the real-time parameter regulation strategy, comprising: When executing the device parameter adjustment instruction, obtain the influent water quality data, effluent water quality data and key node water quality data of the sewage plant; wherein the water quality data in the influent water quality data and effluent water quality data includes but is not limited to PH / T, SS, TP, TN, COD and The key nodes include but are not limited to the biological and reverse tank nodes and the high school sedimentation tank water nodes; the key node water quality data includes but is not limited to DO, ORP, MLSS, TP, TN, COD, Data; The water inlet water condition data is a feedforward signal of the system, and the water outlet water condition data and the key node data are feedback signals of the system; If the water quality data and the standard process difference exceed the preset water quality deviation range, the water quality value is converted into an input variable of a fuzzy control algorithm, and a fuzzy value is output; According to the fuzzy value, a corresponding inference rule is matched in a pre-established fuzzy rule library. If the number of matched inference rules exceeds a single rule threshold, the inference rules are weighted calculated by a fuzzy inference mechanism, and the order of the device parameter adjustment instruction is adjusted.

[0012] Optionally, after the step of matching the real-time water condition fluctuation data with the dynamic parameter control strategy library and the mechanism model to determine the real-time parameter control strategy, and executing the device parameter adjustment instruction according to the real-time parameter control strategy, the step further comprises: After executing the device parameter adjustment instruction, actual carbon emission data is obtained. If the actual carbon emission data is abnormal and the number of continuous abnormal times is equal to an abnormality determination threshold, a carbon emission prediction model parameter updating program is started; According to the carbon emission prediction model parameter updating program, the parameter type of the first prediction module that needs to be corrected is determined; According to the parameter type, the connection weight and bias parameters in the first prediction module are updated to generate a second prediction module; The verification error based on the second prediction module is obtained, and it is judged whether the verification error is abnormal. If not, the carbon emission prediction model is updated according to the second prediction module.

[0013] Optionally, after the step of obtaining the verification error based on the second prediction module and judging whether the verification error is abnormal, if not, updating the carbon emission prediction model according to the second prediction module, the step further comprises: The process unit operating state data applying the updated carbon emission prediction model is obtained; The process unit operating state data is compared with historical data to determine the energy consumption level and emission trend curve of the process unit; If it is detected that the continuous three emission trend curves present an upward state, the corresponding aeration amount adjustment scheme and reflux ratio control scheme are called from the dynamic parameter control strategy library to obtain the process parameter set value of the process unit; The dissolved oxygen concentration fluctuation value based on the process parameter set value is obtained, and the carbon control state is judged according to the dissolved oxygen concentration fluctuation value; If the carbon control state meets the preset carbon control condition, the process parameter set value of the process unit is kept unchanged.

[0014] In addition, to achieve the above object, the embodiment of the present application also provides a sewage treatment plant carbon control system based on artificial intelligence, comprising: A model construction module constructs a carbon emission prediction model according to collected real-time monitoring data, historical data and a mechanism model, wherein the real-time monitoring data comprises water condition fluctuation data and process unit running state data; the water condition fluctuation data comprises water quantity fluctuation data and water quality fluctuation data; the historical data comprises historical water condition fluctuation sequences, historical process unit running parameters and historical carbon emission amounts; An instruction generation module generates a parameter optimization trigger instruction based on processing load data and water condition change rates output by the carbon emission prediction model, and generates a process unit running parameter configuration scheme based on water condition conditions according to the parameter optimization trigger instruction; A strategy construction module acquires a first carbon emission prediction value corresponding to the process unit running parameter configuration scheme, and establishes a dynamic parameter regulation strategy library; An instruction execution module matches real-time water condition fluctuation data with the dynamic parameter regulation strategy library and the mechanism model, determines a real-time parameter regulation strategy, and executes a device parameter adjustment instruction according to the real-time parameter regulation strategy.

[0015] In addition, to achieve the above object, the embodiment of the present application also provides a computer readable storage medium, wherein the computer readable storage medium stores a sewage treatment plant carbon control program based on artificial intelligence, and the sewage treatment plant carbon control program based on artificial intelligence is executed by a processor to realize the method.

[0016] The one or more technical solutions provided in the embodiment of the present application have at least the following technical effects or advantages: (1) The carbon emission prediction model is constructed according to collected real-time monitoring data, historical water condition fluctuation sequences, historical process unit running parameters, historical carbon emission amounts and a mechanism model, wherein the real-time monitoring data comprises water condition fluctuation data and process unit running state data; sewage treatment plant carbon emission is usually estimated by relying on a static model, and the influence of water condition and water quality fluctuation on carbon emission cannot be dynamically reflected, and artificial experience adjustment has hysteresis and is difficult to match real-time working condition changes; by integrating real-time data such as water condition fluctuation data, process unit running parameters and carbon emission, a carbon emission prediction model is constructed through machine learning or a mechanism model, processing load is associated with water condition change rate, and the instantaneous change of carbon emission is quantified; compared with a static model, minute-level prediction of carbon emission can be realized.

[0017] (2) The application generates a parameter optimization trigger instruction based on the processing load data and the water condition change rate output by the carbon emission prediction model, and generates a process unit operation parameter configuration scheme based on the water condition according to the parameter optimization trigger instruction; process parameters are usually fixed at the design value and cannot adapt to water condition fluctuations, resulting in energy waste, while the application automatically generates adjustment instructions based on the load data output by the prediction model, and pre-trains the optimal parameter combination of different water condition intervals through a fuzzy rule base, which can effectively reduce energy consumption and avoid sludge expansion or excessive effluent caused by parameter mismatch.

[0018] (3) The application obtains a first carbon emission prediction value corresponding to the process unit operation parameter configuration scheme, and establishes a dynamic parameter control strategy library; that is, a multi-objective optimization scheme is generated based on the first carbon emission prediction value, and the strategy update period is shortened from "day level" to "minute level".

[0019] (4) The application matches real-time water condition fluctuation data with the dynamic parameter control strategy library to determine a real-time parameter control strategy, and executes a device parameter adjustment instruction according to the real-time parameter control strategy; the real-time strategy is converted into a device control signal and sent to the field device, so that the process unit parameters are adjusted according to the real-time water condition to realize carbon control. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 The flowchart of the first embodiment of the sewage treatment plant carbon control method based on artificial intelligence of the application; Figure 2 The flowchart of the second embodiment of the sewage treatment plant carbon control method based on artificial intelligence of the application; Figure 3 The flowchart of the third embodiment of the sewage treatment plant carbon control method based on artificial intelligence of the application; Figure 4 The terminal structure schematic diagram of the hardware running environment involved in an embodiment scheme of the application. DETAILED DESCRIPTION

[0021] To solve the problem of low energy efficiency and uncontrollable carbon emission of sewage treatment plants caused by water condition fluctuations, the application constructs a dynamic carbon emission prediction model through historical water condition fluctuation data, historical process unit operation parameters, historical carbon emission and mechanism model, and real-time monitoring data, and generates a parameter optimization trigger instruction based on the processing load and water condition change rate output by the model, and then forms a process unit operation parameter configuration scheme matched with the water condition. Through evaluating the carbon emission prediction values of different configuration schemes, a dynamic parameter control strategy library is established, and finally the optimal strategy is matched with real-time water condition fluctuation data, and a device parameter adjustment instruction is automatically executed. The precision and response speed of process control are improved, and the effect of adjusting carbon emission according to different water conditions and reducing energy loss is achieved.

[0022] For a better understanding of the above technical solutions, the exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present application can be more thoroughly understood and the scope of the present application can be fully conveyed to those skilled in the art.

[0023] For a better understanding of the above technical solutions, the exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present application can be more thoroughly understood and the scope of the present application can be fully conveyed to those skilled in the art.

[0024] Embodiment one

[0025] In this embodiment, a sewage treatment plant carbon control method based on artificial intelligence is provided.

[0026] With reference to Figure 1 , the sewage treatment plant carbon control method based on artificial intelligence of the present embodiment comprises the following steps: Step S100: constructing a carbon emission prediction model according to the collected real-time monitoring data, historical data and mechanism model; In this embodiment, a multi-sensor data acquisition system is used to obtain real-time water fluctuation data and process unit operation state data of the sewage treatment plant, as well as historical data. The water fluctuation data includes water fluctuation data and water quality fluctuation data; the historical data includes historical water fluctuation sequence, historical process unit operation parameters and historical carbon emission, the historical data and real-time monitoring data are preprocessed to generate corresponding standardized data sets, then a long short-term memory neural network model is constructed according to the standardized data sets, the network weight parameters are trained through the correlation between historical water fluctuation and carbon emission, a nonlinear mapping relationship between water fluctuation change and carbon emission trend is established, and a trained carbon emission prediction model is obtained. The water fluctuation data and process unit operation parameters at the current time are used as the input of the carbon emission prediction model. This prediction method based on deep learning can accurately identify the influence mode of water fluctuation on carbon emission, greatly improve the prediction accuracy of carbon emission, and provide reliable data support for energy saving and emission reduction decision of the sewage treatment plant.

[0027] As an optional implementation, the real-time monitoring data is preprocessed to generate a multi-dimensional time series data set. The multi-dimensional time series data set contains four dimensions of timestamp, standardized water fluctuation value, process state value and carbon emission concentration value. Arranging the real-time monitoring data into a multi-dimensional time series data set can effectively improve the data quality, provide a reliable data foundation for subsequent intelligent analysis and prediction model, and significantly improve the monitoring accuracy and decision efficiency of the sewage treatment process.

[0028] Exemplarily, the multi-sensor data acquisition interface simultaneously receives real-time data of the flow sensor, the process state monitor and the carbon emission detector of the sewage treatment plant. The sensor precision calibration module corrects the original data by a preset calibration coefficient, for example, the calibration coefficient of the flow sensor is 0.98, and the original reading 15.2 cubic meters per hour is corrected to 14.896 cubic meters per hour, to ensure data accuracy. The calibrated real-time monitoring data are preprocessed by a preset data cleaning rule, and the data cleaning rule includes removing obvious error values, processing data format inconsistencies and the like. Different dimensional monitoring data can be converted into a unified standard format by a standardization parameter conversion module, for example, water level data is converted from cubic meters per hour to a standardized value of 0.76, process unit state is converted from a running state code to a standardized value of 1.0, and carbon emission concentration is converted from milligrams per cubic meter to a standardized value of 0.43, to form a unified time series data dimension matrix, i.e., a standardized data set.

[0029] Optionally, after generating the time series data dimension matrix, abnormal data detection can be performed. An outlier threshold detection algorithm is used to scan the time series data dimension matrix row by row, and upper and lower threshold values are set. When it is detected that a certain data point exceeds the normal range, it is marked as an outlier. For example, when an obviously deviated value appears in the carbon emission concentration data sequence, a cubic spline interpolation method is used to calculate a reasonable replacement value using the normal data points before and after the deviated value, to determine a repaired complete data set. The complete data set is checked according to a processing data integrity verification program, to ensure data continuity and reliability.

[0030] It should be noted that the processing data integrity verification program evaluates data quality by checking data missing rate, outlier proportion and the like. When the data missing rate in a certain time period exceeds 5%, the system triggers a data quality warning, to ensure accuracy of subsequent analysis.

[0031] Optionally, after integrity checking of the complete data set, a timestamp alignment algorithm can be used to synchronize and sort data collected by different sensors according to a unified time reference, to obtain a standardized multi-dimensional time series data set. For example, the flow sensor collects data every 10 seconds, and the carbon emission detector collects data every 30 seconds. An interpolation method is used to unify all data to a time reference of 10-second intervals, to realize data synchronization.

[0032] It should be noted that the collected historical water level fluctuation sequence, historical process unit running parameters and historical carbon emission amount are also subjected to the same preprocessing, to generate a multi-dimensional time series data set corresponding to the historical data, i.e., a standardized data set.

[0033] As another optional implementation, the historical water quality fluctuation sequence is taken as an input feature vector, and the historical carbon emission sequence is taken as a target output value to generate a training data set, wherein each training sample contains water quality data of a preset time length and a corresponding single carbon emission value.

[0034] For example, the historical water quality fluctuation sequence is extracted, a sliding window method is used to take historical water quality fluctuation data of a continuous time period as an input feature vector, and the carbon emission at the corresponding time point of the water quality fluctuation data is taken as a target output value to obtain a complete training data set. The sliding window method constructs a training sample by setting a fixed time window length, for example, 24 hours are selected as the window size, and water quality fluctuation data collected every 10 minutes within 24 hours is used to form a 144-dimensional input feature vector, and the historical carbon emission at the end of the time period is used as a label value. For example, when processing the historical data of a sewage treatment plant for one week, the first sample contains water quality data of the first to 144 time points, and the carbon emission at the 145th time point is predicted, and the second sample contains data of the second to 145 time points, and the carbon emission at the 146th time point is predicted. Through this sliding method, a complete matrix containing thousands of training samples, i.e., a training data set, is finally generated.

[0035] As another optional implementation, after generating the training data set, a loss function value between the predicted carbon emission and the actual carbon emission is calculated based on the training data set by using a back propagation algorithm.

[0036] For example, based on the training data set, a loss function value between the predicted carbon emission and the actual carbon emission is calculated by using a back propagation algorithm, and if the loss function value does not significantly decrease for a plurality of batches, the learning rate parameter is adjusted to determine an optimized network parameter configuration. The back propagation algorithm uses mean square error as an evaluation index when calculating the loss function, and when the loss value of 20 consecutive batches changes by less than 0.001, the learning rate is adjusted from the initial value 0.01 to 0.005 to ensure that the network parameters can continue to be optimized. For example, in a training process, the loss function value gradually decreases from the initial value 2.45 to 0.32 and then stops, at which time the learning rate decay mechanism is triggered to enable the training process to jump out of a local optimal solution and continue to converge.

[0037] As another optional implementation, the network parameter configuration is subjected to prediction accuracy evaluation to determine a first weight parameter, real-time monitoring data is received according to the first weight parameter to construct a prediction input vector, and a carbon emission prediction model is generated in combination with a mechanism model.

[0038] Exemplarily, the network parameter configuration can be calculated by forward propagation, the prediction accuracy can be evaluated by mean square error index, the early stop mechanism is triggered to terminate the training process if the verification error continues to increase, the network training completion state is judged and the optimal weight parameter is saved, the real-time water condition data collected at the current time and the state operation parameters of each process unit are received according to the optimal weight parameter, the prediction input vector is constructed, the forward calculation is performed through the trained long short-term memory neural network, the final carbon emission trend prediction output is obtained, and the construction of the carbon emission prediction model is completed. The early stop mechanism prevents overfitting by monitoring the prediction error on the verification set. When the verification error continues to increase for 15 consecutive training periods, the system immediately terminates the training and saves the current optimal network weight parameter as the first weight parameter. For example, in a certain training, the verification error reaches the minimum value of 0.28 at the 180th period, and then starts to rise. The system triggers the early stop mechanism at the 195th period, avoiding further deterioration of the model performance.

[0039] After obtaining the first weight parameter, the real-time monitored water condition data is integrated based on the first weight parameter obtained by training to form a prediction input vector with time sequence correlation. The vector not only contains conventional water quantity, water quality and other monitoring indicators, but also integrates the theoretical carbon emission benchmark value calculated by the mechanism model and its physical constraint boundary. Then the composite vector is input into the prediction network, and the original prediction result output by the neural network is checked with the theoretical value generated by the mechanism model in both directions: on the one hand, the prediction value that obviously exceeds the abnormal value is corrected by the preset physical rules, such as forcing the negative emission to zero; on the other hand, a dynamic weighting algorithm is used to fuse the theoretical value of the mechanism model and the data-driven prediction result, and finally a carbon emission prediction model is generated which not only conforms to the physical law but also adapts to the actual working condition. The whole process is continuously optimized through real-time feedback mechanism. When the deviation between the monitoring data and the prediction result continuously exceeds the theoretical fluctuation range given by the mechanism model, the online adjustment of the model parameters is automatically triggered to ensure that the prediction system always operates within the physically credible range.

[0040] Optionally, the hidden layer of the long short-term memory neural network contains 128 neural units, which can effectively capture the long-term dependence relationship between water condition fluctuation and carbon emission. In actual prediction of carbon emission, the water condition data sequence of the previous 24 hours at the current time can be received, and the 256-dimensional prediction input vector is constructed by combining the dissolved oxygen concentration of the aeration tank, the sludge return ratio and other process parameters. The trained network model is used for forward calculation to obtain the carbon emission trend prediction value in the next 2 hours.

[0041] Step S200: generating a parameter optimization trigger instruction based on the processing load data and the water condition change rate output by the carbon emission prediction model, and generating a process unit operation parameter configuration scheme based on the water condition according to the parameter optimization trigger instruction; In this embodiment, real-time monitoring data is used as the input vector of the carbon emission prediction model, which can be used to predict the carbon emission under the processing load condition of the future period. If the predicted carbon emission exceeds the preset carbon emission threshold and the water condition change rate exceeds the baseline change range, the parameter optimization program is triggered, and the genetic algorithm is used to search for the optimal device parameter combination to obtain the optimal operation parameter configuration scheme of each process unit under different water conditions. This adaptive parameter optimization mechanism can dynamically adjust the process operation strategy according to the actual water condition fluctuation, effectively balance the treatment effect and carbon emission control target, and provide intelligent decision support for precise operation and management of wastewater treatment plants.

[0042] As an optional implementation, the processing load data and water condition fluctuation information output by the carbon emission prediction model are received. If the carbon emission value in the processing load data exceeds the preset carbon emission threshold, and the water condition change rate exceeds the preset fluctuation amplitude, a parameter optimization trigger instruction is generated.

[0043] Exemplarily, real-time monitoring data is input into the carbon emission prediction model, and processing load data and water condition fluctuation information are output. When the daily processing capacity of a certain wastewater treatment plant reaches 80,000 tons and the carbon emission detection value is 145 tons of carbon dioxide equivalent, it is judged that the carbon emission detection value exceeds the preset threshold of 120 tons. At the same time, the water condition change rate monitoring shows that the current fluctuation amplitude is 35%, which exceeds the normal operation baseline range of 25%. At this time, a parameter optimization trigger instruction is automatically generated to start the subsequent regulation process.

[0044] As another optional implementation, the genetic algorithm program is started according to the parameter optimization trigger instruction, the operation parameter boundary values of the process units are obtained through database query, the population individual matrix is initialized, the population individual matrix includes the device parameter combination of aeration amount, reflux ratio, and sludge concentration, the fitness function is used to receive the device parameter combination, the second carbon emission prediction value and the processing efficiency index of each parameter combination under different water conditions are calculated, if the second carbon emission prediction value is lower than the target threshold, the fitness score of this individual is increased, the new generation population is generated through the cross variation operation of the fitness score, and the fitness evaluation process is repeated until convergence, and the optimal process unit operation parameter configuration scheme of each process unit under different water conditions is obtained.

[0045] Exemplarily, after receiving the trigger instruction, the genetic algorithm program extracts the operation boundary constraints from the process parameter database, including that the aeration amount control range is 2000 to 8000 cubic meters per hour, the sludge return ratio is set to 50% to 200%, and the mixed liquor suspended solid concentration is maintained at 2000 to 6000 mg / L. Based on these constraints, a population matrix containing 100 individuals is initialized, each individual representing a complete set of device parameter combinations. Then the fitness function receives the device parameter combination, and the fitness function is designed to use a multi-objective evaluation mechanism, taking carbon emissions as the main optimization target, while considering treatment efficiency and energy cost. For example, a parameter combination scheme sets the aeration amount to 5500 cubic meters per hour, the return ratio to 120%, and the sludge concentration to 4200 mg / L. After simulation calculation, the carbon emissions under this configuration are 98 tons, which is lower than the target threshold, so a higher fitness score is obtained.

[0046] The fitness score is subjected to crossover and mutation operation, which generates new parameter combinations by simulating the biological evolution process. The top 30% of individuals with the highest fitness scores are selected as parents, and the parameter gene fragments are exchanged through single-point crossover. At the same time, a part of the parameters are randomly mutated with a probability of 5% to ensure population diversity. After 50 generations of continuous evolution iteration, when the optimal solution improvement amplitude is less than 1% for 10 consecutive generations, the algorithm is determined to converge. The final optimization result shows that under low load operation state, the optimal operation parameter configuration scheme is aeration amount 3200 cubic meters per hour, return ratio 85%, sludge concentration 3500 mg / L, and predicted carbon emissions can be reduced to 85 tons. Under high load conditions, the optimal operation parameter configuration scheme is aeration amount 6800 cubic meters per hour, return ratio 150%, and sludge concentration 5200 mg / L, which realizes carbon emission control within 115 tons.

[0047] Step S300: obtaining a first carbon emission prediction value corresponding to the process unit operation parameter configuration scheme, and establishing a dynamic parameter control strategy library; In this embodiment, the first carbon emission prediction value corresponding to the process unit operation parameter configuration scheme and the water quality fluctuation characteristic data are obtained, a dynamic parameter control strategy library is established, a strategy index mechanism is constructed through carbon emission threshold correlation analysis of the parameter configuration scheme, and if the predicted carbon emissions corresponding to the parameter configuration scheme are in the critical threshold position, it is marked as a high-priority control strategy, and a hierarchical management parameter control strategy library is obtained.

[0048] As an optional implementation, the optimal process unit operation parameter configuration scheme output by the genetic algorithm is received, the corresponding first carbon emission prediction value and water condition fluctuation characteristic data are obtained through database query, if the difference between the first carbon emission prediction value and the preset carbon emission threshold value is less than the critical range, the parameter configuration scheme is marked as a high priority category, and the parameter configuration data of the high priority category is clustered to calculate the carbon emission threshold correlation degree of each configuration scheme.

[0049] Exemplarily, after receiving the process unit operation parameter configuration scheme output by the genetic algorithm, the corresponding carbon emission prediction value and water condition fluctuation characteristic data are obtained through a database query mechanism. When a configuration scheme shows that the aeration amount is 4500 cubic meters per hour and the reflux ratio is 110%, the query obtains that the carbon emission prediction value is 118 tons of carbon dioxide equivalent, and the difference from the preset carbon emission threshold value of 120 tons is only 2 tons, which is less than the critical range of 5 tons. The configuration scheme is marked as a high priority category. The parameter configuration data of the high priority category needs to be clustered to identify the similarity pattern. A clustering algorithm based on Euclidean distance is used, and key parameters such as aeration amount, reflux ratio, and sludge concentration are used as feature vectors for grouping. For example, when it is identified that the aeration amounts of three configuration schemes are 4200, 4500, and 4800 cubic meters per hour respectively, and the reflux ratios are all within the interval of 100% to 120%, these parameter configuration schemes are grouped into the same cluster.

[0050] For the parameter configuration schemes in each cluster, the carbon emission threshold correlation degree is further calculated, which is determined by the ratio of the actual carbon emission prediction value to the preset carbon emission threshold value. Specifically, when the carbon emission prediction value of a configuration scheme is 118 tons and the preset carbon emission threshold value is 120 tons, the correlation degree is 98.3%. This index can directly reflect the optimization potential of each configuration scheme under the condition of meeting the carbon emission constraint, and the higher the correlation degree, the greater the margin from the upper limit of the threshold, and the more significant the optimization value.

[0051] Optionally, each cluster can be sorted according to the calculated correlation degree, and parameter configuration schemes with higher correlation degrees are preferentially selected as candidate objects for process optimization implementation, while invalid schemes with negative correlation degrees are automatically excluded, thereby forming the final operation parameter configuration scheme.

[0052] As another optional implementation, when it is detected that the carbon emission threshold correlation degree exceeds a preset correlation reference value, an emergency control index is assigned, a hierarchical index structure strategy classification result is obtained, a storage structure of a dynamic parameter control strategy library is established through the strategy classification result, a strategy retrieval mechanism is constructed according to the emergency control index, the calling weight of each control scheme is determined, a hierarchical management rule is generated according to the calling weight, and then a dynamic parameter control strategy library is generated.

[0053] Exemplarily, when the carbon emission prediction value of a certain configuration scheme is 116 tons, the corresponding carbon emission threshold correlation degree is calculated to be 96.7%, which exceeds the preset correlation reference value of 85%, and thus an emergency control index mark is obtained. This marking mechanism can ensure the rapid response capability in the critical state. The preset correlation reference value is a critical point obtained by analyzing historical accident data. The response mechanism adopts a three-level index architecture. The first level index corresponds to the configuration scheme with the carbon emission value in the range of (115 tons, 120 tons]. The second level index covers the interval of (110 tons, 115 tons]. The third level index includes the scheme below 110 tons. Each level is further subdivided according to the water quality fluctuation characteristics, forming a multi-dimensional strategy index system. The storage structure of the dynamic parameter control strategy library adopts an association table design, which maps the process unit operation parameter configuration, carbon emission prediction value, water quality fluctuation mode and calling weight. When the water quality change rate is 30% and the carbon emission risk is high, the corresponding configuration scheme is assigned a calling weight of 0.9, while the calling weight is set to 0.3 in the case of low risk. The hierarchical management rule determines the strategy priority according to the calling weight. The weight threshold can be set to 0.7. The process unit operation parameter configuration scheme that exceeds the weight threshold has the qualification of automatic calling, and the construction of the dynamic parameter control strategy library is completed.

[0054] Optionally, after the dynamic parameter control strategy library is constructed, its actual application process exhibits a complete closed-loop control system. When the water quality change rate is monitored to reach 32%, a multi-dimensional collaborative response is started. First, the fast retrieval function of the strategy library is activated. Through the pre-set hash index structure, the mode matching of the massive stored strategies is completed within milliseconds, and the parameter configuration group most suitable for the current working condition is accurately located. The weight coefficients of each candidate strategy are comprehensively evaluated. When the calling weight of the configuration scheme exceeds the weight threshold of 0.7, it is determined to use the configuration scheme "aeration amount 5200 cubic meters per hour, reflux ratio 125%", and automatically execute the control instruction. The actual effect is continuously monitored, the actual carbon emission change is recorded, and the weight of the configuration scheme is updated according to the change result. If the emission reduction meets the standard, the weight increases by 0.1.

[0055] Step S400: match the real-time water quality fluctuation data with the dynamic parameter control strategy library and the mechanism model, determine the real-time parameter control strategy, and execute the equipment parameter adjustment instruction according to the real-time parameter control strategy.

[0056] In this embodiment, real-time water condition fluctuation data is matched with a dynamic parameter regulation strategy library and a mechanism model to determine a real-time regulation strategy. If the current water condition change rate exceeds a preset change threshold and the carbon emission prediction value of the matched strategy is lower than a safety threshold, the corresponding device parameter adjustment instruction is automatically executed to obtain real-time updated device operation parameters. This closed-loop control mechanism realizes complete automation of the monitoring, decision-making, and execution processes, significantly improving the response speed and control accuracy of the sewage treatment system to water condition fluctuations, while ensuring the effective achievement of environmental protection goals through carbon emission prediction constraints.

[0057] As an optional implementation, a fluctuation feature vector of real-time water condition fluctuation data is obtained, the fluctuation feature vector is input into a mechanism model for theoretical carbon emission pre-calculation to obtain the physical feasible region boundary under the current working condition. At the same time, the fluctuation feature vector is matched with the dynamic parameter regulation strategy library in terms of similarity, the matching result needs to be verified for physical feasibility by the mechanism model, and the deviation between the actual carbon emission and the predicted carbon emission is continuously monitored during the execution process. When the deviation exceeds the theoretical capacity given by the mechanism model, the dynamic adjustment of the strategy and the online update of the strategy library are triggered to form a closed-loop control flow of "monitoring-matching-verification-execution-feedback". Then, according to the matching result that has passed the feasibility verification, the parameter regulation strategy record is extracted, the device operation parameters in the parameter regulation strategy record are calculated for carbon emission prediction, and if the predicted carbon emission value is less than a preset safety threshold, the device parameter adjustment instruction is extracted from the parameter regulation strategy record to determine the real-time parameter regulation strategy. The fluctuation feature vector contains multi-dimensional characteristic parameters such as flow rate change rate, water level fluctuation amplitude, and time interval, forming a standardized data structure for subsequent matching analysis.

[0058] For example, the sensor network data acquisition mechanism obtains the operation state information of the sewage treatment system in real time through distributed monitoring nodes. When the water flow sensor detects that the current flow rate is 2.8 cubic meters per second, and the previous recorded value is 2.3 cubic meters per second, the difference is calculated as 0.5 cubic meters per second. Since the difference exceeds the preset change threshold of 0.3 cubic meters per second, the fluctuation feature vector generation program is triggered. The fluctuation feature vector is obtained, and the cosine similarity algorithm is used for strategy library retrieval of the fluctuation feature vector. When a feature vector containing a flow rate change rate of 27%, a water level rise of 0.5 meters, and a time interval of 15 minutes is received, the historical records in the strategy library are traversed to identify a matching item with a similarity of 0.87, which exceeds the preset matching reference value of 0.8. The corresponding strategy record is extracted, which contains key parameter information such as adjustment of the aeration device power to 75% and setting of the sludge return pump high frequency to 45 Hz.

[0059] The mechanism model verification module receives the matching result, calculates the physical feasible region boundary under the current working condition through the hydrodynamic model, and confirms that the strategy parameter does not exceed the theoretical capacity range. During execution, the carbon emission monitoring system continuously compares the actual carbon emission with the predicted value, and when the actual carbon emission deviates from the predicted value by 4% (without exceeding the 8% theoretical capacity given by the mechanism model), the current strategy is continued to be executed.

[0060] The neural network carbon emission prediction module receives the device operating parameters in the strategy record as input features. When the aeration power is 75% and the reflux pump frequency is 45 Hz, the multi-layer perceptron network calculates and predicts the carbon emission value to be 102 tons of carbon dioxide equivalent. Since the predicted value is lower than the preset safety threshold of 110 tons, the feasibility of the parameter adjustment scheme is confirmed, the parameter control strategy is determined, and the specific device parameter adjustment instructions are extracted from the strategy record.

[0061] As another optional implementation, after determining the parameter control strategy, the specific device parameter adjustment instructions are extracted from the strategy record, and the target device control interface address is determined according to the parameter type identifier in the adjustment instructions. The parameter modification command is sent to the target device through the industrial communication protocol, and the device response state is obtained. If the device response state is successful, the modified device operating parameters are written into the parameter configuration database.

[0062] Exemplarily, the parameter type identifier parsing mechanism determines the control interface address according to the device code in the adjustment instructions. When the aeration device identifier AER_001 is identified, the corresponding Modbus communication address 192.168.1.15 and port number 502 are obtained through the device mapping table query. The system uses the industrial Ethernet protocol to build a communication connection, sends a power adjustment command to the target device, and modifies the current 75% power to 68%. The device controller returns an acknowledgement signal after receiving the command, and the response status code is displayed as 0x01, indicating that the parameter modification is successfully executed. The parameter configuration database update mechanism ensures real-time synchronization of the system operating state. When the device response state is confirmed to be successful, the system writes the modified operating parameters into the corresponding record in the configuration database. The power parameter of the aeration device AER_001 is updated from 75% to 68%, and the modification timestamp and operation source identifier are recorded.

[0063] In the embodiment, by fusing real-time monitoring data, historical operation data and mechanism model, a carbon emission prediction model is constructed, effectively solving the problem of delayed response of regulation and control caused by post-emission monitoring in the related art. Meanwhile, a parameter optimization triggering mechanism based on water quality change characteristics is established, solving the problem that the fixed parameter mode is difficult to adapt to water quality fluctuations. The constructed strategy library breaks through the randomness and limitations of manual experience parameter adjustment through intelligent matching algorithm, shortening the decision-making time of process adjustment from hours to seconds. The dynamic parameter regulation strategy library has a continuous self-learning ability and can be continuously optimized with the accumulation of operation data, which can provide a replicable intelligent solution for low-carbon operation in the field of wastewater treatment. The "prediction-decision-execution-optimization" control mode can effectively predict the carbon emissions under different water conditions and reduce energy consumption.

[0064] Embodiment Two

[0065] Based on Embodiment One, another embodiment of the present application is proposed, referring to Figure 2 After matching the real-time water quality fluctuation data with the dynamic parameter regulation strategy library and the mechanism model, determining the real-time parameter regulation strategy, and executing the equipment parameter adjustment instruction according to the real-time parameter regulation strategy, the following steps are included: Step S500: When executing the equipment parameter adjustment instruction, acquiring the influent water quality data, effluent water quality data and key node water quality data of the wastewater treatment plant; wherein the water quality data in the influent water quality data and effluent water quality data includes but is not limited to PH / T, SS, TP, TN, COD and The key nodes include but are not limited to the biological reaction tank node and the high-speed sedimentation tank water node; the key node water quality data includes but is not limited to DO, ORP, MLSS, TP, TN, COD, The influent water quality data is the feedforward signal of the system, and the effluent water quality data and the key node data are the feedback signals of the system; Step S600: If the difference between the water quality data and the standard process exceeds the preset water quality deviation range, convert the water quality data into the input variable of the fuzzy control algorithm, and output the fuzzified data; In the embodiment, the fuzzy control algorithm can be used to finely regulate the equipment parameters of each process unit, and the uncertain factors in the parameter adjustment process can be processed through the fuzzy inference mechanism. If multiple process units are adjusted at the same time, the control instruction is executed according to the preset priority order to determine the final equipment operation state.

[0066] ​As an optional implementation, when executing the device parameter adjustment instruction, the full-process water quality data of the sewage plant is collected in real time by the distributed sensor network, including the water quality and quantity parameters of the inlet, outlet and key process nodes (such as biochemical reaction tank and high-efficiency sedimentation tank). The inlet water quality data (PH / T, SS, TP, TN, COD, , etc.) are input as feedforward signals into the control model to predict process disturbance and adjust device parameters in advance; the outlet water quality and key node data (DO, ORP, MLSS, TP, TN, COD, , etc.) are used as feedback signals to check process effect in real time and trigger dynamic correction. When the deviation of the inlet or node water quality parameters (such as COD, ) from the standard process value exceeds the preset threshold (such as COD deviation > 15%, deviation > 10%), the exceeding parameters are automatically converted into input variables of the fuzzy control algorithm, which are fuzzified into “low”, “medium” and “high” three-level language variables through the membership function, and the fuzzy adjustment instructions (such as “aeration quantity increases by 20%” and “return ratio decreases by 10%”) are output in combination with the process rule base, and finally the accurate device control parameters are obtained by defuzzification.

[0067] For example, when a sewage plant executes the instruction “aeration device power adjustment to 75%”, it simultaneously starts full-process water quality monitoring, first collects data, the inlet sensor feedbacks the current PH value of 7.2 (standard range 6.8-7.5), COD concentration of 320 mg / L (standard value ≤ 250 mg / L, deviation + 28%), and concentration of 35 mg / L (standard value ≤ 25 mg / L, deviation + 40%), showing that the inlet water quality has a significant fluctuation; the dissolved oxygen (DO) concentration in the biochemical reaction tank is 1.8 mg / L (standard value 2-4 mg / L, deviation -25%), and the mixed liquor suspended solids (MLSS) concentration is 3800 mg / L (standard value 3000-4500 mg / L, within the normal range); the outlet sensor shows that the COD concentration is 45 mg / L (standard value ≤ 50 mg / L), and concentration is 4.2 mg / L (standard value ≤ 5 mg / L), and the outlet water quality has not exceeded the standard. Since the deviations of the inlet COD and concentrations from the standard process value exceed the preset threshold (15% and 10% respectively), the system automatically triggers the fuzzy control logic, taking the inlet COD (320 mg / L) and (35 mg / L) as input variables, which are fuzzified into “low”, “medium” and “high” three-level language variables through the trapezoidal membership function—among which the COD is determined as “high” (membership degree 0.8) due to the deviation + 28%, The deviation of +40% was judged as "high" (membership degree 0.9). This is consistent with the process rule library's statement that "if the influent COD is high and..." Following the rule of "increasing aeration rate and sludge return ratio when the aeration rate is high," the system outputs fuzzy adjustment commands: "Aeration rate increase = 0.7 (high), sludge return ratio adjustment = 0.5 (medium)." Precise control parameters are obtained through defuzzification using the center-of-gravity method: aeration equipment power is increased from 75% to 85%, and sludge return pump frequency is increased from 45Hz to 50Hz. After 10 minutes of adjustment, the DO concentration in the biological reactor rises to 2.5 mg / L, and the effluent COD concentration drops to 38 mg / L. When the concentration dropped to 3.5 mg / L, all parameters returned to the standard range, verifying the dynamic correction capability of fuzzy control for water quality fluctuations.

[0068] Step S700: Match the corresponding inference rule in the pre-established fuzzy rule library according to the fuzzy value. If the number of matched inference rules exceeds the threshold of a single rule, the inference rules are weighted and calculated through the fuzzy inference mechanism to adjust the order of the device parameter adjustment instructions.

[0069] As an optional implementation, the corresponding inference rule is matched in a pre-established fuzzy rule base according to the fuzzy value. If the number of matched rules exceeds the threshold of a single rule, the fuzzy inference mechanism is used to perform weighted calculation on multiple rules, and the parameter adjustment instruction is obtained through defuzzification.

[0070] For example, the fuzzy rule base matching mechanism retrieves the corresponding inference rules based on the fuzzy values. When a fuzzy state of "low temperature" and "medium pressure" is identified, three relevant rules are matched: Rule A suggests increasing the heater power by 15%, Rule B suggests increasing the circulation pump frequency by 8 Hz, and Rule C suggests adjusting the inlet valve opening to 70%. Since the number of matched rules is 3, exceeding the single rule threshold of 2, the fuzzy inference mechanism is activated for weighted calculation. The weighted calculation process allocates weights based on the confidence level of each rule. The confidence level of Rule A is 0.8, the confidence level of Rule B is 0.6, and the confidence level of Rule C is 0.7. Through centroid defuzzification, a comprehensive control command is finally obtained, which adjusts the heater power to 78%, sets the circulation pump frequency to 52 Hz, and adjusts the inlet valve opening to 68%.

[0071] As another optional embodiment, when there are multiple parameter adjustment instructions, a priority sorting algorithm can be used to sort the parameter adjustment instructions according to the importance of the process units. If multiple process units need to be adjusted at the same time and there is a resource conflict, the control instructions are executed in the preset priority order, the specific parameter modification values of each device are obtained, the parameter modification values are sent to the process unit through the device control interface, if the device response time exceeds the preset response time limit, the device abnormal information is recorded and the standby control scheme is switched to, and the actual running state parameters of each process unit device are determined.

[0072] Exemplarily, the priority sorting algorithm sorts the control instructions according to the importance of the process units. When the anaerobic tank, the aerobic tank and the secondary sedimentation tank need parameter adjustment at the same time, the anaerobic tank is set as priority 1, the aerobic tank is set as priority 2, and the secondary sedimentation tank is set as priority 3 according to the influence degree of the treatment effect. If there is a conflict in the heating device resources, the temperature control instruction of the anaerobic tank is executed first, and then the control requirements of other process units are processed in turn. The specific parameter modification values of each device are obtained, and the parameter modification values are sent to the process unit through the device control interface. The device control interface communication mechanism ensures the reliable transmission of the parameter modification instruction. When the power adjustment instruction is sent to the anaerobic tank heater, the response time limit is set to 3 seconds. If the device does not return an acknowledgement signal within the specified time, the abnormal information of the device AH_003 is recorded, and the standby control scheme is switched to, and the temperature adjustment target is achieved through the auxiliary heating device. This fault-tolerant mechanism ensures the continuity and reliability of the process parameter control, significantly improves the automation control level and operation stability of the sewage treatment system.

[0073] In this embodiment, after the parameter control instruction is sent to the target device, the pressure value and the temperature value of each process unit are continuously monitored, and a closed-loop fuzzy control system is constructed to realize intelligent dynamic optimization of the sewage treatment process parameters. When the device parameter adjustment causes the temperature to exceed the standard, the real-time data is automatically converted into fuzzy variables, the order and parameters of the adjustment instruction are dynamically optimized through rule base matching and multi-rule weighted calculation, thereby significantly improving the process stability and reducing energy consumption.

[0074] Embodiment three

[0075] Based on embodiment one, another embodiment of the present application is proposed, referring to Figure 3 After the step of matching the real-time water condition fluctuation data with the dynamic parameter control strategy library, determining the real-time parameter control strategy, and executing the device parameter adjustment instruction according to the real-time parameter control strategy, the following steps are included: Step S800: After executing the device parameter adjustment instruction, actual carbon emission data is obtained, and if the actual carbon emission data is abnormal and the number of continuous abnormal times is equal to an abnormality determination threshold, a carbon emission prediction model parameter updating procedure is started; Step S900: According to the carbon emission prediction model parameter updating procedure, the type of the parameter of the first prediction module that needs to be corrected is determined. In this embodiment, after the device parameter adjustment is performed, the actual carbon emission data after the parameter adjustment is obtained, and a comparison analysis is performed on the actual carbon emission data and the carbon emission prediction value output by the carbon emission prediction model. If the deviation between the actual carbon emission data and the first carbon emission prediction value exceeds a permissible range, the carbon emission prediction model parameter is updated, and the carbon emission prediction accuracy is corrected.

[0076] As an optional implementation, the concentration values of nitrous oxide and methane during the operation of the process unit can be collected by a carbon emission monitoring sensor. If the collection interval time of the concentration values exceeds a preset monitoring frequency, a data compensation mechanism is triggered to perform interpolation calculation on the emission data of the actual time period to obtain a complete actual carbon emission monitoring data sequence. According to the actual carbon emission monitoring data sequence, a pre-trained neural network prediction module is called. If there is a significant difference between the process parameters of the current time period and the parameter range of the historical training data, the input weight of the prediction module is dynamically adjusted to obtain a carbon emission prediction value corresponding to the time period. A deviation calculation formula is used to perform difference operation on the actual carbon emission data and the carbon emission prediction value. If the absolute value of the calculated deviation exceeds a preset permissible range and the number of continuous occurrences reaches an abnormality determination threshold, a model parameter updating procedure is started.

[0077] Illustratively, the anaerobic tank is configured with methane and nitrous oxide concentration sensors, and the sampling frequency is set to collect data every 30 seconds. Under normal operating conditions, the methane concentration is maintained within the range of 800-1200 ppm, and the nitrous oxide concentration is 50-150 ppm. During the monitoring process from 14:30 to 14:33 on a certain day, the system detects that the methane concentration data at 14:32 is missing, and immediately triggers the data compensation mechanism: based on the linear interpolation algorithm, the compensation value of 14:32 is calculated to be 1000 ppm based on the measured value of 980 ppm at 14:30 and the measured value of 1020 ppm at 14:33, thereby forming a complete time series data set.

[0078] The neural network prediction module is called, and the neural network prediction module adopts a multi-layer perceptron structure. The input layer includes 12 process parameter nodes such as temperature, pH value and dissolved oxygen. When it is identified that the temperature of the current anaerobic tank is 28 degrees Celsius, and the temperature range of the historical training data is 30 to 38 degrees Celsius, the neural network prediction module automatically adjusts the input weight of the temperature parameter from 0.8 to 0.6 to adapt to the current working condition. After processing by the activation function in the hidden layer, the output layer generates the methane concentration of 1050 ppm and the nitrous oxide concentration of 110 ppm in the next 2 hours.

[0079] The difference calculation formula is used for difference operation, and the deviation calculation mechanism compares the actual monitoring value with the prediction value in real time. The actual monitoring value (methane 1020 ppm, nitrous oxide 115 ppm) at 14:33 is compared with the prediction value: the absolute value of the methane deviation is |1020-1050|=30ppm (not exceeding the preset tolerance 45ppm), and the absolute value of the nitrous oxide deviation is |115-110|=5ppm (normal). However, the methane deviation of the subsequent 5 consecutive sampling points (14:34-14:36) continues to expand, with a maximum deviation of 65ppm, triggering the abnormal judgment threshold. The model parameter updating program is started. At this time, it is identified that the types of parameters that need to be corrected include the hidden layer weight matrix and the output layer bias vector.

[0080] As another optional implementation, according to the carbon emission prediction model parameter updating program, the types of parameters of the first prediction module that need to be corrected are determined.

[0081] Exemplarily, when the prediction deviation exceeds the standard for 5 times in a row, the parameter tracing analysis is started: the contribution of each network layer to the total error is calculated through the back propagation algorithm, it is found that the sensitivity of the hidden layer weight matrix to abnormal input of temperature rises to 42%, and the output layer bias vector shows a systematic deviation in error transmission, with a deviation of more than 2 times the standard deviation. At this time, combined with the real-time working condition data and model structure analysis, it is identified that these two types of parameters have produced adaptive degradation due to long-term non-adjustment, such as a decrease of 37% in the L2 norm of the weight matrix and a deviation of 15% in the bias vector distribution from the training period benchmark, so as to accurately lock the hidden layer weight matrix and the output layer bias vector as the core parameters to be corrected. This triple verification mechanism based on error contribution analysis, parameter health evaluation and working condition adaptability detection ensures the pertinence and effectiveness of parameter correction.

[0082] Step S1000: updating the connection weight and bias parameter in the first prediction module according to the parameter type, to generate a second prediction module; As an optional implementation, the connection weight and bias parameter in the first prediction module can be iteratively updated by a gradient descent optimization algorithm to generate a second prediction module.

[0083] Illustratively, the gradient descent optimization algorithm adjusts the network parameters through a backpropagation mechanism, with a learning rate set to 0.01 and a batch size of 32 samples. In the first iteration, the key connection weight of the third hidden layer is adjusted from the initial value of 0.45 to 0.52 along the error gradient direction, while the corresponding bias parameter is optimized from 0.12 to 0.08. This precise parameter update significantly improves the model's performance on the validation set: for the validation data containing temperature fluctuations, the average prediction error is reduced from 38 ppm for the first prediction module to 24 ppm for the second prediction module, a decrease of 37%. The parameter-optimized prediction module can more accurately identify the impact of process operation mode changes on carbon emissions. When the aeration rate of the aerobic tank is adjusted from 180 cubic meters per hour to 220 cubic meters per hour, the corrected model predicts that the carbon dioxide emissions will peak within 30 minutes and then gradually fall to a stable level. This dynamic prediction capability provides reliable decision support for operators, effectively avoiding abnormal fluctuations in carbon emissions caused by process adjustments, and significantly improving the environmental management and control level and operational efficiency of wastewater treatment.

[0084] Step S1100: Obtain the validation error based on the second prediction module, and determine whether the validation error is abnormal. If not, update the carbon emission prediction model according to the second prediction module.

[0085] As an optional implementation, the validation error refers to the error indicator calculated when the performance of the second prediction module is evaluated using an independent validation dataset after updating the model parameters. The validation dataset contains historical temperature, pH, dissolved oxygen, and other input features and corresponding true carbon dioxide concentration values. If the updated validation error decreases compared to the error value before updating, i.e., there is no abnormality, save the updated model parameter configuration to obtain the corrected carbon emission prediction accuracy indicator.

[0086] Exemplarily, when it is detected that the carbon dioxide concentration prediction deviation exceeds 45 ppm for 5 times in succession, the parameter updating procedure is started. Assuming that the mean square error of the model before updating on the validation set is 80 ppm2, after 3 rounds of iterative adjustment of the hidden layer weights and output layer bias by the gradient descent algorithm, the same validation set is retested, and it is found that the mean square error is reduced to 65 ppm2, with a reduction of 18.75%. It is determined that this parameter update is effective, and the new weight matrix and bias vector are immediately saved, and the updated prediction accuracy index is recorded. For example, in the actual operation of an anaerobic tank, the updated model improves the average deviation of the predicted value of carbon emission in the next 2 hours from the actual monitoring value from ±58 ppm to ±43 ppm, significantly improving the early warning reliability. This dynamic updating mechanism based on the comparison of validation errors not only avoids the performance fluctuations of the model caused by local optimization, but also ensures that each parameter adjustment brings quantifiable precision improvement.

[0087] As another optional implementation, after updating the carbon emission prediction model according to the second prediction module, the energy consumption level and emission trend of each process unit can be recalculated. The process unit operating state data of the updated carbon emission prediction model is obtained, the process unit operating state data is compared with the historical data, the energy consumption level and the emission trend curve of the process unit are determined, the energy consumption level and the emission trend curve are read by the feedback control program, and if it is detected that the emission trend curves of three process units in succession present an upward state, the corresponding aeration amount adjustment scheme and reflux ratio control scheme are retrieved from the dynamic parameter control strategy library, and the updated process parameter set value is obtained.

[0088] Exemplarily, the energy consumption level calculation mechanism recalculates the power of each process unit based on the corrected prediction accuracy. The biochemical tank power calculation involves comprehensive energy consumption analysis of aeration blowers, mixers, and circulating pumps. When the aeration blower power of the aerobic tank in the secondary sewage treatment A2O process reaches 85 kilowatts, while the historical average value of the same period is 72 kilowatts, the unit is marked as a high energy consumption state. At the same time, the gas emission rate calculation shows that the methane emission during this period is 12 cubic meters per hour, which is about 30% higher than the normal working condition. The generation of the emission trend curve uses a sliding window algorithm for data processing. Continuous monitoring of the emission data of the anaerobic tank, anoxic tank, and aerobic tank, when the carbon dioxide emission rates of the three units are detected to increase from 8 cubic meters per hour, 6 cubic meters per hour, and 15 cubic meters per hour to 11 cubic meters per hour, 9 cubic meters per hour, and 19 cubic meters per hour respectively, the feedback control program immediately triggers the parameter regulation mechanism. The dynamic parameter regulation strategy library contains preset schemes for different working conditions. When the emission trend shows an upward trend, the aeration amount reduction adjustment scheme is called to gradually reduce the aeration amount of the aerobic tank from 220 cubic meters per hour to 180 cubic meters per hour. The internal reflux ratio is adjusted from 200% to 150%, and the external reflux ratio is adjusted from 80% to 60%. This coordinated control strategy can effectively balance the biochemical reaction efficiency and energy consumption control.

[0089] As a further optional implementation, after obtaining the process parameter set value, a real-time monitoring program is used to apply the process parameter set value, to obtain a dissolved oxygen concentration fluctuation amplitude value in the running process after parameter adjustment, and to determine a carbon control state according to the dissolved oxygen concentration fluctuation value; if the carbon control state meets a preset carbon control condition, the process parameter set value of the process unit is kept unchanged.

[0090] Exemplarily, the real-time monitoring program verifies and evaluates the parameter adjustment effect. The dissolved oxygen concentration monitoring shows that the adjusted value is stable in the range of 2.5 mg / L to 3.2 mg / L, with a fluctuation amplitude of only 0.7 mg / L, which is much smaller than the preset threshold of 1.2 mg / L. The sludge settling ratio is kept in the ideal interval of 25% to 30%, and it is determined that the current running stability index meets the requirements, a confirmation signal of stable running state is obtained, and the current aeration amount set and reflux ratio control parameters are kept unchanged. At this time, the aeration amount is maintained at 180 cubic meters per hour, and the internal reflux ratio is fixed at 150%.

[0091] Optionally, after obtaining the confirmation signal of stable running state, monitoring is continued, and if it is monitored that the influent water quality changes significantly or the treatment load fluctuates, the parameter regulation process is restarted. For example, when the chemical oxygen demand COD concentration suddenly rises from 300 mg / L to 450 mg / L, or the treatment water condition increases from 800 cubic meters per hour to 1000 cubic meters per hour, the load fluctuation detection module identifies the working condition change and automatically restarts the parameter regulation process.

[0092] In the embodiment, after the device parameter adjustment is performed, the sewage treatment process is continuously monitored, and the related parameters of the carbon emission prediction model are optimized according to the monitoring result, and the closed-loop control mechanism realizes the dynamic balance of energy consumption optimization and emission control. By adjusting the process parameters in real time, the overall energy consumption can be reduced under the premise of meeting the effluent quality standards, and the emission of greenhouse gases can be reduced.

[0093] Embodiment four

[0094] In the embodiment of the present application, an artificial intelligence-based carbon control device for a sewage treatment plant is provided.

[0095] Reference Figure 4 , Figure 4 The terminal structure diagram of the hardware running environment involved in an embodiment of the present application is shown in the figure.

[0096] As Figure 4 shown, the control terminal can include a processor 1001, such as a CPU, a network interface 1003, a memory 1004, and a communication bus 1002. The communication bus 1002 is used to realize the connection and communication between these components. The network interface 1003 can optionally include a standard wired interface, a wireless interface (such as a WI-FI interface). The memory 1004 can be a high-speed RAM memory, or a stable memory (non-volatile memory), such as a magnetic disk memory. The memory 1004 can optionally be a storage device independent of the aforementioned processor 1001.

[0097] Those skilled in the art can understand that Figure 4 the terminal structure shown in the figure does not constitute a limitation on the terminal, and can include more or fewer components than the figure, or combine certain components, or different component arrangements.

[0098] As Figure 4 shown, the memory 1004, as a computer storage medium, can include an operating system, a network communication module, and an artificial intelligence sewage treatment plant carbon control program.

[0099] In Figure 4 the hardware structure of the artificial intelligence sewage treatment plant carbon control device, the processor 1001 can call the artificial intelligence sewage treatment plant carbon control program stored in the memory 1004, and perform the following operations: According to the collected real-time monitoring data, historical data and mechanism model, a carbon emission prediction model is constructed, wherein the real-time monitoring data includes water condition fluctuation data and process unit running state data; the water condition fluctuation data includes water quantity fluctuation data and water quality fluctuation data; the historical data includes historical water condition fluctuation sequence, historical process unit running parameter and historical carbon emission; Based on the processing load data and water condition change rate output by the carbon emission prediction model, a parameter optimization trigger instruction is generated, and a process unit running parameter configuration scheme based on water condition is generated according to the parameter optimization trigger instruction; A first carbon emission prediction value corresponding to the process unit running parameter configuration scheme is obtained, and a dynamic parameter control strategy library is established; The real-time water condition fluctuation data is matched with the dynamic parameter control strategy library and the mechanism model to determine a real-time parameter control strategy, and a device parameter adjustment instruction is executed according to the real-time parameter control strategy.

[0100] Optionally, the processor 1001 can call the sewage treatment plant carbon control program of artificial intelligence stored in the memory 1004, and further perform the following operations: If the carbon emission value in the processing load data exceeds the preset carbon emission threshold, and the water condition change rate exceeds the preset fluctuation amplitude, the parameter optimization trigger instruction is generated; According to the parameter optimization trigger instruction, a genetic algorithm program is started, running parameter boundary values of the process unit are obtained, and a population individual matrix of the genetic algorithm is initialized according to the running parameter boundary values, the population individual matrix containing device parameter combinations of aeration quantity, reflux ratio and sludge concentration; Based on the water condition, a second carbon emission prediction value and a processing efficiency index of the device parameter combination are evaluated, and a fitness score is generated; If the fitness score does not meet the expected condition, the device parameter combination is subjected to cross mutation operation, and the device parameter combination after cross mutation is reevaluated to update the fitness score; When the fitness score meets the expected condition, the device parameter combination corresponding to the expected condition is output, and the process unit running parameter configuration scheme is generated.

[0101] Optionally, the processor 1001 can call the sewage treatment plant carbon control program of artificial intelligence stored in the memory 1004, and further perform the following operations: The historical water condition fluctuation sequence is taken as an input feature vector, and the historical carbon emission sequence is taken as a target output value to generate a training data set, wherein each training sample contains water condition data and a single carbon emission value corresponding to a preset time length; calculating, based on the training data set, a loss function value between a predicted carbon emission and an actual carbon emission through a back propagation algorithm; adjusting a network parameter configuration according to the loss function value, and evaluating a prediction accuracy of the network parameter configuration to determine a first weight parameter; receiving the real-time monitoring data according to the first weight parameter, constructing a prediction input vector, and generating the carbon emission prediction model in combination with the mechanism model.

[0102] Optionally, the processor 1001 can invoke the artificial intelligence sewage treatment plant carbon management program stored in the memory 1004, and further perform the following operations: If the difference between the first carbon emission prediction value and the preset carbon emission threshold value is less than a critical range, a carbon emission threshold value correlation degree of the process unit operation parameter configuration scheme is calculated; If the carbon emission threshold value correlation degree exceeds a preset correlation reference value, an emergency control index identifier is assigned, and a hierarchical index structure strategy classification result is generated; A storage structure of a dynamic parameter control strategy library is established through the strategy classification result, a strategy retrieval mechanism is constructed according to the emergency control index identifier, and a calling weight of the process unit operation parameter configuration scheme control scheme is determined; The dynamic parameter control strategy library is generated according to the calling weight.

[0103] Optionally, the processor 1001 can invoke the artificial intelligence sewage treatment plant carbon management program stored in the memory 1004, and further perform the following operations: Obtain a fluctuation feature vector of the water condition fluctuation data, perform similarity matching between the fluctuation feature vector, the dynamic parameter control strategy library and the mechanism model, and extract a parameter control strategy record according to the matching result; Perform carbon emission prediction calculation on the equipment operation parameters in the parameter control strategy record, and if the predicted carbon emission value is less than a preset safety threshold value, extract an equipment parameter adjustment instruction from the parameter control strategy record to determine the real-time parameter control strategy; Determine a target equipment control interface address according to the parameter type identifier in the real-time parameter control strategy, and send the equipment parameter adjustment instruction to the target equipment.

[0104] Optionally, the processor 1001 can invoke the artificial intelligence sewage treatment plant carbon management program stored in the memory 1004, and further perform the following operations: When the equipment parameter adjustment instruction is executed, obtain the influent water condition data, the effluent water condition data and the key node water condition data of the sewage treatment plant; wherein the water quality data in the influent water condition data and the effluent water condition data includes but is not limited to PH / T, SS, TP, TN, COD and numerical values; the key nodes include but are not limited to the anoxic tank node and the high-rate settling tank node; the key node water quality data include but are not limited to DO, ORP, MLSS, TP, TN, COD, data of the data; The water inflow condition data is a feedforward signal of the system, and the water outflow condition data and the key node data are feedback signals of the system; If the water quality data deviates from the standard process by more than a preset water quality deviation range, the water quality numerical value is converted into an input variable of a fuzzy control algorithm, and a fuzzified numerical value is output; According to the fuzzified numerical value, a corresponding inference rule is matched in a pre-established fuzzy rule library. If the number of matched inference rules exceeds a single rule threshold, the inference rules are weighted calculated by a fuzzy inference mechanism, and the order of the device parameter adjustment instruction is adjusted.

[0105] Optionally, the processor 1001 can call the artificial intelligence sewage treatment plant carbon control program stored in the memory 1004, and further perform the following operations: After executing the device parameter adjustment instruction, actual carbon emission data is obtained. If the actual carbon emission data is abnormal and the number of continuous abnormal times is equal to an abnormality determination threshold, a carbon emission prediction model parameter updating program is started; According to the carbon emission prediction model parameter updating program, the parameter type of the first prediction module that needs to be corrected is determined; According to the parameter type, the connection weight and bias parameters in the first prediction module are updated to generate a second prediction module; The verification error based on the second prediction module is obtained, and it is judged whether the verification error is abnormal. If not, the carbon emission prediction model is updated according to the second prediction module.

[0106] Optionally, the processor 1001 can call the artificial intelligence sewage treatment plant carbon control program stored in the memory 1004, and further perform the following operations: The process unit operating state data applying the updated carbon emission prediction model is obtained; The process unit operating state data is compared with historical data to determine the energy consumption level and emission trend curve of the process unit; If it is detected that the continuous three emission trend curves present an upward state, the corresponding aeration amount adjustment scheme and reflux ratio control scheme are retrieved from the dynamic parameter regulation strategy library to obtain the process parameter set value of the process unit; The dissolved oxygen concentration fluctuation value based on the process parameter set value is obtained, and the carbon control state is judged according to the dissolved oxygen concentration fluctuation value; If the carbon control state meets a preset carbon control condition, the process parameter setting value of the process unit is kept unchanged.

[0107] In addition, to achieve the above object, the embodiment of the present application further provides a sewage treatment plant carbon control system based on artificial intelligence, which comprises: A model construction module is configured to construct a carbon emission prediction model according to collected real-time monitoring data, historical data and a mechanism model, wherein the real-time monitoring data comprises water condition fluctuation data and process unit operation state data; the water condition fluctuation data comprises water quantity fluctuation data and water quality fluctuation data; the historical data comprises historical water condition fluctuation sequences, historical process unit operation parameters and historical carbon emission amounts; An instruction generation module is configured to generate a parameter optimization trigger instruction based on processing load data and water condition change rate output by the carbon emission prediction model, and generate a process unit operation parameter configuration scheme based on water condition conditions according to the parameter optimization trigger instruction; A strategy construction module is configured to obtain a first carbon emission prediction value corresponding to the process unit operation parameter configuration scheme, and establish a dynamic parameter regulation strategy library; An instruction execution module is configured to match real-time water condition fluctuation data with the dynamic parameter regulation strategy library and the mechanism model, determine a real-time parameter regulation strategy, and execute a device parameter adjustment instruction according to the real-time parameter regulation strategy.

[0108] In addition, to achieve the above object, the embodiment of the present application further provides a computer readable storage medium, which stores a sewage treatment plant carbon control program based on artificial intelligence, and the sewage treatment plant carbon control program based on artificial intelligence is executed by a processor to realize the sewage treatment plant carbon control method based on artificial intelligence.

[0109] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can adopt a completely hardware embodiment, a completely software embodiment or an embodiment combining software and hardware aspects. Moreover, the present application can adopt a computer program product in the form of one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program codes.

[0110] The present application is described in reference to the flow diagrams and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the present application. It will be understood that each block of the flow diagrams and / or block diagrams, and combinations of blocks in the flow diagrams and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flow diagrams and / or block diagrams block or blocks. Figure 1 The flow diagrams and / or block diagrams in the flow diagrams and / or block diagrams illustrate the architecture, functionality, and operation of possible implementations of apparatuses and computer program products according to various embodiments of the present application. In this regard, each block in the flow diagrams and / or block diagrams can represent a module, segment, or portion of code, which comprises one or more executable Figure 1 The flow diagrams and / or block diagrams in the flow diagrams and / or block diagrams illustrate the architecture, functionality, and operation of possible implementations of apparatuses and computer program products according to various embodiments of the present application. In this regard, each block in the flow diagrams and / or block diagrams can represent a module, segment, or portion of code, which comprises one or more executable

[0111] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the Figure 1 The flow diagrams and / or block diagrams in the flow diagrams and / or block diagrams illustrate the architecture, functionality, and operation of possible implementations of apparatuses and computer program products according to various embodiments of the present application. In this regard, each block in the flow diagrams and / or block diagrams can represent a module, segment, or portion of code, which comprises one or more executable Figure 1 The flow diagrams and / or block diagrams in the flow diagrams and / or block diagrams illustrate the architecture, functionality, and operation of possible implementations of apparatuses and computer program products according to various embodiments of the present application. In this regard, each block in the flow diagrams and / or block diagrams can represent a module, segment, or portion of code, which comprises one or more executable

[0112] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flow diagrams and / or block diagrams block or blocks. Figure 1 The flow diagrams and / or block diagrams in the flow diagrams and / or block diagrams illustrate the architecture, functionality, and operation of possible implementations of apparatuses and computer program products according to various embodiments of the present application. In this regard, each block in the flow diagrams and / or block diagrams can represent a module, segment, or portion of code, which comprises one or more executable Figure 1 The flow diagrams and / or block diagrams in the flow diagrams and / or block diagrams illustrate the architecture, functionality, and operation of possible implementations of apparatuses and computer program products according to various embodiments of the present application. In this regard, each block in the flow diagrams and / or block diagrams can represent a module, segment, or portion of code, which comprises one or more executable

[0113] It should be noted that any references made herein to elements or components of the application are intended to represent exemplary embodiments of the application and are intended to provide a platform from which further embodiments can be derived. It should also be noted that, in the claims, any reference signs placed between parentheses to the right of a comma do not constitute restrictions on the scope of the claims. Use of the word "a" or "an" does not exclude a plurality. The suffix "(s)" is used in the singular sense only when recited directly after the noun to which it is appended. The application can be implemented by means of both hardware and software, and any combination of hardware and software. Several of the appended claims utilize "means for" or "step for" clauses. These clauses are intended to perform a limitation with respect to the elements named in such clauses. The use of such clauses is to be interpreted as written limitation provided that any of the claimed elements can also be directly or indirectly connected to other claim elements excluding any electrical connector there between.

[0114] Although preferred embodiments of the application have been described, those skilled in the art will recognize that additional modifications and variations can be made thereto. It is therefore intended to include all such modifications and variations as fall within the scope of the application.

[0115] Obviously, many modifications and variations of the present application are possible in light of the above teachings. It is, therefore, to be understood that within the scope of the appended claims and their equivalents, the application can be practiced otherwise than as specifically described.

Claims

1. An artificial intelligence-based carbon management method for a sewage treatment plant, characterized by, The application relates to a carbon emission prediction method and device. The carbon emission prediction model is constructed based on real-time monitoring data, historical data and a mechanism model, wherein the real-time monitoring data comprises water condition fluctuation data and process unit running state data; the water condition fluctuation data comprises water quantity fluctuation data and water quality fluctuation data; the historical data comprises a historical water condition fluctuation sequence, historical process unit running parameters and historical carbon emission quantity; a parameter optimization trigger instruction is generated based on processing load data and water condition change rate output by the carbon emission prediction model, and a process unit running parameter configuration scheme based on water condition is generated according to the parameter optimization trigger instruction; a first carbon emission prediction value corresponding to the process unit running parameter configuration scheme is obtained, and a dynamic parameter regulation strategy library is established; real-time water condition fluctuation data is matched with the dynamic parameter regulation strategy library and the mechanism model, a real-time parameter regulation strategy is determined, and a device parameter adjustment instruction is executed according to the real-time parameter regulation strategy. The step of generating the parameter optimization trigger instruction based on the processing load data and the water condition change rate output by the carbon emission prediction model and generating the process unit running parameter configuration scheme based on the water condition comprises the following steps: If the carbon emission quantity value in the processing load data exceeds a preset carbon emission threshold value and the water condition change rate exceeds a preset fluctuation amplitude, the parameter optimization trigger instruction is generated; A genetic algorithm program is started according to the parameter optimization trigger instruction, running parameter boundary values of the process unit are obtained, a population individual matrix of the genetic algorithm is initialized according to the running parameter boundary values, the population individual matrix comprises a device parameter combination of aeration quantity, reflux ratio and sludge concentration; a second carbon emission prediction value and a processing efficiency index of the device parameter combination are evaluated based on the water condition, and an adaptability score is generated; 2. The artificial intelligence based carbon management method for wastewater treatment plant as claimed in claim 1 wherein, If the adaptability score does not satisfy an expected condition, the device parameter combination is subjected to a cross variation operation, the device parameter combination after the cross variation operation is reevaluated, and the adaptability score is updated; When the adaptability score satisfies the expected condition, a device parameter combination satisfying the expected condition is output, and the process unit running parameter configuration scheme is generated. The step of constructing the carbon emission prediction model based on the real-time monitoring data, the historical data and the mechanism model comprises the following steps: The historical water condition fluctuation sequence is taken as an input feature vector, and the historical carbon emission quantity sequence is taken as a target output value, so that a training data set is generated, wherein each training sample comprises water condition data and a single carbon emission value corresponding to a preset time length; A loss function value between a predicted carbon emission quantity and an actual carbon emission quantity is calculated by a back propagation algorithm based on the training data set; Network parameter configuration is adjusted according to the loss function value, the prediction precision of the network parameter configuration is evaluated, and a first weight parameter is determined; 3.The artificial intelligence-based carbon management method for a wastewater treatment plant of claim 1, wherein, The real-time monitoring data is received according to the first weight parameter, a prediction input vector is constructed, and the carbon emission prediction model is generated in combination with the mechanism model. ​ ​ ​ ​ 4.The artificial intelligence-based carbon management method for a wastewater treatment plant of claim 1, wherein The step of obtaining the first carbon emission prediction value corresponding to the process unit operation parameter configuration scheme, and establishing a dynamic parameter regulation strategy library comprises: If the difference between the first carbon emission prediction value and the preset carbon emission threshold value is less than a critical range, then calculate the carbon emission threshold value correlation degree of the process unit operation parameter configuration scheme; If the carbon emission threshold value correlation degree exceeds a preset correlation reference value, then assign an emergency regulation index identifier, and generate a hierarchical index structure strategy classification result; Establish a storage structure of the dynamic parameter regulation strategy library through the strategy classification result, and construct a strategy retrieval mechanism according to the emergency regulation index identifier, to determine the calling weight of the process unit operation parameter configuration scheme regulation scheme; Generate a hierarchical management dynamic parameter regulation strategy library according to the calling weight. 5.The artificial intelligence-based carbon management method for a wastewater treatment plant of claim 1, wherein, The step of matching the real-time water quality fluctuation data with the dynamic parameter regulation strategy library and the mechanism model, determining a real-time parameter regulation strategy, and executing a device parameter adjustment instruction according to the real-time parameter regulation strategy comprises: Obtain a fluctuation feature vector of the water quality fluctuation data, perform similarity matching of the fluctuation feature vector with the dynamic parameter regulation strategy library and the mechanism model, and extract a parameter regulation strategy record according to the matching result; Perform carbon emission prediction calculation on the device operation parameters in the parameter regulation strategy record, if the predicted carbon emission value is less than a preset safety threshold value, then extract a device parameter adjustment instruction from the parameter regulation strategy record, and determine the real-time parameter regulation strategy; Determine a target device control interface address according to the parameter type identifier in the real-time parameter regulation strategy, and send the device parameter adjustment instruction to the target device. 6.The artificial intelligence-based carbon management method for a wastewater treatment plant of claim 1, wherein, After the step of matching the real-time water quality fluctuation data with the dynamic parameter regulation strategy library and the mechanism model, determining a real-time parameter regulation strategy, and executing a device parameter adjustment instruction according to the real-time parameter regulation strategy, the following steps are included: In executing the device parameter adjustment instruction, water inflow condition data, water outflow condition data and key node water condition data of the sewage plant are acquired; wherein the water quality data in the water inflow condition data and the water outflow condition data includes but is not limited to PH / T, SS, TP, TN, COD and The numerical value; the key node includes but is not limited to the biological reaction tank node and the high school sedimentation tank water node; the key node water quality data includes but is not limited to DO, ORP, MLSS, TP, TN, COD, The data. The inlet water quality data is a feedforward signal of the system, and the outlet water quality data and the key node data are feedback signals of the system; If the difference between the water quality data and the standard process exceeds a preset water quality deviation range, then convert the water quality value into an input variable of a fuzzy control algorithm, and output a fuzzified value; According to the fuzzified value, match corresponding inference rules in a pre-established fuzzy rule library, if the number of matched inference rules exceeds a single rule threshold value, then perform weighted calculation on the inference rules through a fuzzy inference mechanism, and adjust the order of the device parameter adjustment instruction. 7.The artificial intelligence-based carbon management method for a wastewater treatment plant of claim 1, wherein, After the step of matching the real-time water quality fluctuation data with the dynamic parameter regulation strategy library and the mechanism model, determining a real-time parameter regulation strategy, and executing a device parameter adjustment instruction according to the real-time parameter regulation strategy, the following steps are included: After executing the device parameter adjustment instruction, obtain actual carbon emission data, if the actual carbon emission data is abnormal and the number of continuous abnormal times is equal to an abnormality judgment threshold value, then start a carbon emission prediction model parameter updating program; According to the carbon emission prediction model parameter updating procedure, the type of the parameters of the first prediction module that needs to be corrected is determined; According to the type of the parameters, the connection weight and bias parameters in the first prediction module are updated to generate a second prediction module; Obtain the validation error based on the second prediction module, and determine whether the validation error is abnormal. If not, update the carbon emission prediction model according to the second prediction module.

8. The artificial intelligence based carbon management method for wastewater treatment plant as claimed in claim 7 wherein, The step of obtaining the validation error based on the second prediction module, and determining whether the validation error is abnormal, if not, updating the carbon emission prediction model according to the second prediction module, is followed by: Obtain the process unit operating state data applying the updated carbon emission prediction model; Compare the process unit operating state data with historical data to determine the energy consumption level and emission trend curve of the process unit; If it is detected that the continuous three emission trend curves present an upward state, the corresponding aeration amount adjustment scheme and reflux ratio control scheme are retrieved from the dynamic parameter regulation strategy library to obtain the process parameter set value of the process unit; Obtain the dissolved oxygen concentration fluctuation value based on the process parameter set value, and determine the carbon control state according to the dissolved oxygen concentration fluctuation value; If the carbon control state meets the preset carbon control condition, the process parameter set value of the process unit is kept unchanged.

9. An artificial intelligence-based carbon management system for a wastewater treatment plant, characterized in that, It includes: A model construction module constructs a carbon emission prediction model according to collected real-time monitoring data, historical data and mechanism models, wherein the real-time monitoring data includes water fluctuation data, process unit operating state data; the water fluctuation data includes water quantity fluctuation data and water quality fluctuation data; the historical data includes historical water fluctuation sequence, historical process unit operating parameters and historical carbon emission; An instruction generation module generates a parameter optimization trigger instruction based on the processing load data and water fluctuation rate output by the carbon emission prediction model, and generates a process unit operating parameter configuration scheme based on water condition according to the parameter optimization trigger instruction; A strategy construction module acquires a first carbon emission prediction value corresponding to the process unit operating parameter configuration scheme, and establishes a dynamic parameter regulation strategy library; An instruction execution module matches real-time water fluctuation data with the dynamic parameter regulation strategy library and the mechanism model to determine a real-time parameter regulation strategy, and executes a device parameter adjustment instruction according to the real-time parameter regulation strategy.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores an artificial intelligence-based carbon control program for sewage treatment plants, which is executed by a processor to implement the method of any one of claims 1-8. The computer readable storage medium stores an artificial intelligence-based carbon control program for sewage treatment plants, which is executed by a processor to implement the method of any one of claims 1-8.

Images