CNN-BiGRU-based sewage treatment agent feeding concentration control method

Abstract

The invention relates to the field of sewage treatment agent feeding, and discloses a CNN-BiGRU-based sewage treatment agent feeding concentration control method, which comprises the following steps: S10, collecting data of sewage treatment agent feeding concentration; s20, carrying out preprocessing on the collected data; s30, performing feature extraction on the data by using a convolutional neural network; s40, taking the features extracted by the convolutional neural network as an input sequence, and inputting the input sequence into a bidirectional gating circulation unit for sequence modeling; s50, according to the trained model, performing drug delivery concentration prediction on new input data; and S60, in combination with the data monitored in real time, a feedback control strategy is adopted to adjust the prediction result, and the control strategy is updated in real time. Sequence modeling is carried out by using the convolutional neural network and the bidirectional gating circulation unit, so that the characteristics in the sewage treatment process can be extracted more accurately, and the concentration of the put medicament can be predicted. Therefore, the agent feeding strategy can be optimized, and the sewage treatment efficiency and quality can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of sewage treatment agent placement, and in particular to a sewage treatment agent placement concentration control method based on CNN-BiGRU. Background Art

[0002] In the sewage treatment process, the addition of chemicals is one of the key links and has an important impact on the treatment effect and system stability.

[0003] Commonly used methods include quantitative delivery and fixed ratio delivery. These methods often rely on manual experience and trial and error, and have the following shortcomings:

[0004] Lack of precise control of chemical dosage: Traditional methods often cannot accurately predict and control the dosage concentration of chemical. Due to the lack of accurate modeling of complex water quality and environmental factors, chemical dosage often results in excessive or insufficient dosage, which affects the treatment effect and resource utilization efficiency.

[0005] High manual operation and management costs: Traditional methods require a lot of manual operation and management, including data collection, adjustment of dosage concentration, and monitoring of effects. This not only increases operating costs, but is also easily affected by human factors, resulting in unstable treatment effects;

[0006] Lack of real-time monitoring and feedback control: Traditional methods can only adjust the release at fixed time intervals, and cannot monitor and control the release concentration in real time. This results in the treatment system being unable to respond to emergencies or changes in a timely manner, affecting the treatment effect;

[0007] Waste of resources and environmental risks: Due to the lack of precise dosing control, excessive dosage of chemicals is often used, resulting in waste of resources and environmental risks. At the same time, due to the inability to adjust the dosage concentration in time, the treatment system may not be able to cope with changes in water quality and environmental factors, further increasing waste of resources and environmental risks.

[0008] Based on the above problems, a sewage treatment agent concentration control method based on CNN-BiGRU is proposed to improve the problems in the existing technology. Summary of the invention

[0009] In view of the shortcomings of the prior art, the present invention provides a sewage treatment agent dosage concentration control method based on CNN-BiGRU, which solves the problems of poor treatment effect and resource utilization efficiency of the existing sewage treatment concentration control method.

[0010] To achieve the above objectives, the present invention is implemented by the following technical scheme: a method for controlling the concentration of sewage treatment reagents based on CNN-BiGRU, comprising the following steps:

[0011] S10. Collect data on the concentration of sewage treatment agents;

[0012] S20, preprocessing the collected data;

[0013] S30, using convolutional neural network (CNN) to extract features from data;

[0014] S40, taking the features extracted by the convolutional neural network as the input sequence and inputting it into a bidirectional gated recurrent unit (BiGRU) for sequence modeling;

[0015] S50, predicting the concentration of the drug delivery for new input data according to the trained model;

[0016] S60, combining the real-time monitoring data, using a feedback control strategy to adjust the prediction results, and updating the control strategy in real time;

[0017] S70, monitor the dosage concentration in real time, and control the dosage concentration according to the prediction result.

[0018] Preferably, the pretreatment step in step S20 comprises the following steps:

[0019] S21, processing missing values, using one or more of the above calculations including interpolation, mean filling or similarity-based methods;

[0020] S22, outlier processing, detect and process outliers, and use statistical methods for calculation;

[0021] S23, feature normalization, use normalization processing to eliminate the dimensional differences of different features.

[0022] Preferably, in step S30, the convolutional neural network (CNN) includes multiple parallel convolutional layers and pooling layers.

[0023] Preferably, in step S40, the bidirectional gated recurrent unit (BiGRU) includes multiple layers of forward GRU and backward GRU.

[0024] Preferably, in step S50, the model training and verification uses the root mean square error (RMSE) as the loss function, and the stochastic gradient descent algorithm is used to update the model parameters; the root mean square error function expression is:

[0025]

[0026] Where (L) represents the root mean square error, (N) represents the number of samples, and (y i ) represents the actual observed value of the (i)th sample, represents the model's predicted value for the (i)th sample;

[0027] The function expression of the stochastic gradient descent algorithm is:

[0028]

[0029] Among them, (θ) represents the parameters of the model, (α) represents the learning rate (used to control the step size of parameter update), and (J(θ)) represents the gradient of the loss function with respect to the parameter (θ).

[0030] Preferably, in step S10, the concentration data includes inlet flow rate, outlet water quality, and temperature parameters.

[0031] Preferably, in step S60, the real-time monitoring data includes the concentration of sewage treatment agents, influent water quality, and meteorological data.

[0032] Preferably, the feedback control strategy in step S60 uses the weighted sum of the model prediction error and the controller gain as the correction amount for the drug delivery concentration.

[0033] Preferably, in step S50, the model training uses a cross-validation method to evaluate the model and select the optimal model parameters in the evaluation.

[0034] Preferably, in step S60, the feedback control strategy is based on the real-time monitored dosage concentration and system status, and the control target is achieved by adjusting the dosage concentration of the agent, and the control is performed using a PID control algorithm.

[0035] The present invention provides a method for controlling the concentration of sewage treatment reagents based on CNN-BiGRU. It has the following beneficial effects:

[0036] 1. The present invention uses convolutional neural networks (CNN) and bidirectional gated recurrent units (BiGRU) for sequence modeling, which can more accurately extract the characteristics of the sewage treatment process and predict the concentration of the dosage agent. This can optimize the dosage agent strategy and improve the efficiency and quality of sewage treatment.

[0037] 2. The present invention can reduce the waste and over-dosage of drugs and improve the efficiency of resource utilization through precise drug delivery control. At the same time, according to real-time monitoring data and model prediction results, adjustments can be made according to actual needs to avoid over-processing or under-processing, further improving the effective utilization of resources.

[0038] 3. The present invention uses advanced deep learning algorithms of convolutional neural networks and bidirectional gated recurrent units to achieve automated and intelligent control of the sewage treatment process. By training models and optimizing control strategies, the concentration of reagents can be automatically adjusted, reducing the need for manual intervention and operation, and improving the operating efficiency and stability of the sewage treatment system.

[0039] 4. The present invention monitors the key parameters of the concentration of the feed, the quality of the influent water and the meteorological data in real time, combines the model prediction error and the controller gain, uses the feedback control strategy to adjust the prediction results, and updates the control strategy in real time. In this way, the deviation of the feed concentration can be discovered and corrected in time to ensure the stable operation of the sewage treatment system. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a schematic diagram of the process steps of the present invention;

[0041] Figure 2 Schematic diagram of the pretreatment process steps of the present invention. DETAILED DESCRIPTION

[0042] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] Example:

[0044] Please refer to the attached Figure 1 The embodiment of the present invention provides a method for controlling the concentration of sewage treatment reagents based on CNN-BiGRU, comprising the following steps:

[0045] S10, collect data related to the concentration of sewage treatment agents: In the S10 stage, it is necessary to collect data related to the concentration of sewage treatment agents. These data can include various parameters in the sewage treatment system, such as influent flow, chemical content of sewage, and environmental factors. At the same time, it is also necessary to record the concentration of agents and the corresponding treatment effect. Data collection can be carried out through sensors, monitoring equipment or manual collection.

[0046] S20, preprocess the collected data: In the S20 stage, the collected data needs to be preprocessed to prepare for model training and prediction. Preprocessing includes data cleaning, removing outliers, filling missing values, data smoothing and normalization steps. Cleaning data can remove invalid or erroneous data points, filling missing values ​​can use interpolation or other methods, smoothing data can reduce the impact of noise, and normalization can map the data to a suitable range.

[0047] S30, use convolutional neural network (CNN) to extract features from data; In the S30 stage, convolutional neural network (CNN) can be used to extract features from preprocessed data. Convolutional neural network (CNN) can effectively capture the spatial and temporal correlations in the data, thereby extracting useful features. By constructing appropriate convolutional layers and pooling layers, convolutional neural network (CNN) can automatically learn local patterns and global features in the data.

[0048] S40: The features extracted from the convolutional neural network are used as the input sequence and input into the BiGRU for sequence modeling. In the S40 stage, the features extracted from the convolutional neural network (CNN) are used as the input sequence and input into the BiGRU for sequence modeling. The BiGRU can effectively capture long-term dependencies in sequence data and provide a global understanding of historical information. Through the forward and backward calculation process, the BiGRU can consider both past and future information at the same time, thereby better modeling the sequence.

[0049] S50: According to the trained model, the new input data is used to predict the concentration of the drug. In the S50 stage, according to the trained model, the new input data can be used to predict the concentration of the drug. By inputting the input data into the trained model, the model can give the corresponding prediction result of the drug concentration according to the learned features and sequence modeling capabilities.

[0050] S60, combined with real-time monitoring data, the prediction results are adjusted using feedback control strategies, and the control strategies are updated in real time; In the S60 stage, the prediction results can be adjusted using feedback control strategies combined with real-time monitoring data, and the control strategies are updated in real time. By comparing with the real-time monitoring data, the difference between the actual drug delivery concentration and the prediction can be obtained, and adjustments can be made based on the difference. According to the feedback control principle, the difference can be reduced by adjusting the control strategy, making the prediction results more accurate and stable.

[0051] S70, monitor the dosage concentration in real time and control the dosage concentration according to the prediction results. In the S70 stage, it is necessary to monitor the dosage concentration in real time and control the dosage concentration according to the prediction results. By monitoring the dosage concentration in real time, the actual treatment effect can be obtained and compared with the prediction results. According to the comparison results, the dosage concentration can be adjusted to achieve better sewage treatment effect. This process can be repeated to achieve real-time control and optimization of the dosage concentration.

[0052] The calculation function expression of the bidirectional gated recurrent unit (BiGRU) is:

[0053] First, define the forward calculation process of BiGRU:

[0054]

[0055] in, represents the forward hidden state at time (t), represents the input at time (t), Represents the forward hidden state at time (t-1);

[0056] Then, define the backward calculation process of BiGRU:

[0057]

[0058] in, represents the backward hidden state at time (t), represents the input at time (t), Represents the backward hidden state at time (t+1);

[0059] Finally, the forward and backward hidden states are concatenated to get the output of the BiGRU:

[0060]

[0061] Among them, (h t ) represents the BiGRU output at time (t). Through the forward and backward calculation process of BiGRU, the input sequence can be modeled and the contextual information in the sequence can be captured.

[0062] Please refer to the attached Figure 2 , the preprocessing step in step S20 includes the following steps:

[0063] S21, processing missing values, using one or more of the above calculations including interpolation, mean filling or similarity-based methods;

[0064] S22, outlier processing, detect and process outliers, and use statistical methods for calculation;

[0065] S23, feature normalization, use normalization processing to eliminate the dimensional differences of different features.

[0066] In step S30, the convolutional neural network (CNN) includes multiple parallel convolutional layers and pooling layers; the convolutional layer extracts local features of the input data through convolution operations, and then performs nonlinear transformations through activation functions. The pooling layer is used for downsampling to reduce the spatial dimensions of the feature map while retaining important feature information. By stacking multiple convolutional layers and pooling layers, CNN can gradually extract high-level features of the data.

[0067] In step S40, the bidirectional gated recurrent unit (BiGRU) includes multiple layers of forward GRU and backward GRU; GRU is a gated recurrent unit that controls the flow and retention of information through a gating mechanism. The forward GRU calculates the hidden state from the beginning of the sequence, while the backward GRU calculates the hidden state from the end of the sequence. Through the forward and backward calculation process, BiGRU can consider past and future information at the same time and provide a global understanding of historical information.

[0068] In step S50, the root mean square error (RMSE) is used as the loss function for model training and verification, and the stochastic gradient descent algorithm is used to update the model parameters; the root mean square error function expression is:

[0069]

[0070] Where (L) represents the root mean square error, (N) represents the number of samples, and (y i ) represents the actual observed value of the (i)th sample, represents the model's predicted value for the (i)th sample;

[0071] The function expression of the stochastic gradient descent algorithm is:

[0072]

[0073] Among them, (θ) represents the parameters of the model, (α) represents the learning rate (used to control the step size of parameter update), and (J(θ)) represents the gradient of the loss function with respect to the parameter (θ).

[0074] In step S10, the concentration data includes inlet flow, outlet water quality, and temperature parameters; inlet flow refers to the flow rate entering the sewage treatment system, outlet water quality refers to the water quality characteristics of the treated sewage, and temperature parameters refer to the temperature information of the sewage. These data can be used to build models and make predictions.

[0075] In step S60, the real-time monitoring data includes the concentration of sewage treatment agents, influent water quality, and meteorological data; the concentration of sewage treatment agents is used to monitor the concentration of agents in real time, the influent water quality is used to understand the water quality before treatment, and the meteorological data is used to understand the impact of environmental factors on the sewage treatment effect.

[0076] In step S60, the feedback control strategy uses the weighted sum of the model prediction error and the controller gain as the correction amount for the drug delivery concentration; the model prediction error refers to the difference between the predicted value and the actual value, and the controller gain is a gain parameter set according to system requirements and performance requirements. Through the weighted sum operation, the correction amount can be calculated to adjust the drug delivery concentration to achieve better control effect.

[0077] In step S50, the model training uses the cross-validation method to evaluate the model and select the optimal model parameters in the evaluation; cross-validation is a commonly used model evaluation method that divides the data set into a training set and a validation set, and evaluates the performance of the model through multiple training and validation. Based on the evaluation results, the model parameters with the best performance can be selected.

[0078] In step S60, the feedback control strategy is based on the real-time monitored dosage concentration and system status, and achieves the control target by adjusting the dosage concentration of the drug, and uses the PID control algorithm for control; the PID control algorithm calculates the control signal based on the current error (the difference between the actual value and the target value), the cumulative error (the integral of the error) and the error change rate (the differential of the error) to achieve dynamic adjustment of the dosage concentration of the drug.

[0079] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A sewage treatment agent concentration control method based on CNN-BiGRU, characterized in that: The following steps are involved: S10. Collect data on the concentration of sewage treatment agents; S20, preprocessing the collected data; S30, using convolutional neural network (CNN) to extract features from data; S40, taking the features extracted by the convolutional neural network as the input sequence and inputting it into a bidirectional gated recurrent unit (BiGRU) for sequence modeling; S50, predicting the concentration of the drug delivery for new input data according to the trained model; S60, combining the real-time monitoring data, using a feedback control strategy to adjust the prediction results, and updating the control strategy in real time; S70, monitor the dosage concentration in real time, and control the dosage concentration according to the prediction result.

2. A method for controlling the concentration of sewage treatment reagents based on CNN-BiGRU according to claim 1, characterized in that: The pre-processing step in step S20 comprises the following steps: S21, processing missing values, using one or more of the above calculations including interpolation, mean filling or similarity-based methods; S22, outlier processing, detect and process outliers, and use statistical methods for calculation; S23, feature normalization, use normalization processing to eliminate the dimensional differences of different features.

3. A method for controlling the concentration of sewage treatment reagents based on CNN-BiGRU according to claim 1, characterized in that: In the step S30, the convolutional neural network (CNN) includes multiple parallel convolutional layers and pooling layers.

4. A method for controlling the concentration of sewage treatment reagents based on CNN-BiGRU according to claim 1, characterized in that: In the step S40, the bidirectional gated recurrent unit (BiGRU) includes multiple layers of forward GRU and backward GRU.

5. The method for controlling the concentration of sewage treatment reagents based on CNN-BiGRU according to claim 1, characterized in that: In the step S50, the model training and verification uses the root mean square error (RMSE) as the loss function, and the stochastic gradient descent algorithm is used to update the model parameters; the root mean square error function expression is: Where (L) represents the root mean square error, (N) represents the number of samples, and (y i ) represents the actual observed value of the (i)th sample, represents the model's predicted value for the (i)th sample; The function expression of the stochastic gradient descent algorithm is: Among them, (θ) represents the parameters of the model, (α) represents the learning rate (used to control the step size of parameter update), and (J(θ)) represents the gradient of the loss function with respect to the parameter (θ).

6. The method for controlling the concentration of sewage treatment reagents based on CNN-BiGRU according to claim 1, characterized in that: In the step S10, the concentration data includes inlet flow rate, outlet water quality, and temperature parameters.

7. The method for controlling the concentration of sewage treatment reagents based on CNN-BiGRU according to claim 1, characterized in that: In the step S60, the real-time monitoring data includes the concentration of sewage treatment agents, the quality of influent water, and meteorological data.

8. The method for controlling the concentration of sewage treatment reagents based on CNN-BiGRU according to claim 1, characterized in that: In the step S60, the feedback control strategy uses the weighted sum of the model prediction error and the controller gain as the correction amount for the drug delivery concentration.

9. The method for controlling the concentration of sewage treatment reagents based on CNN-BiGRU according to claim 1, characterized in that: In step S50, the model training uses a cross-validation method to evaluate the model and select the optimal model parameters in the evaluation.

10. The method for controlling the concentration of sewage treatment reagents based on CNN-BiGRU according to claim 1, characterized in that: In the step S60, the feedback control strategy is based on the real-time monitored dosage concentration and system status, and the control target is achieved by adjusting the dosage concentration of the agent, and the PID control algorithm is used for control.

Images