A method and device for water quality prediction and optimal control of desulfurization wastewater of a thermal power plant

Abstract

The present application belongs to the field of industrial process automatic control and wastewater treatment technology, and discloses a water quality prediction and optimization control method and device for desulfurization wastewater of a thermal power plant, comprising: obtaining real-time data of process variables of a target desulfurization system; pre-processing and feature selection are performed on the real-time data of process variables of the target desulfurization system to obtain real-time data of features strongly related to chloride ion concentration; based on the real-time data of features related to chloride ion concentration and in combination with a pre-constructed chloride ion concentration prediction model, a real-time prediction value of chloride ion concentration is obtained; based on the real-time prediction value of chloride ion concentration and in combination with a pre-set control logic, an optimization control is performed on a chemical feeding device in the target desulfurization system; the present application realizes accurate online prediction of chloride ion concentration in desulfurization wastewater, and realizes closed-loop optimization control of chemical feeding and discharge based on high-precision prediction value, effectively reduces wastewater discharge and chemical consumption, and realizes safe, economic and environmentally-friendly collaborative operation of the desulfurization system.

Description

Technical Field

[0001] This invention belongs to the field of industrial process automatic control and wastewater treatment technology, and specifically relates to a method and device for water quality prediction and optimization control of desulfurization wastewater from thermal power plants. Background Technology

[0002] In the field of coal-fired power generation, limestone-gypsum wet flue gas desulfurization technology has become the mainstream desulfurization process widely used in coal-fired power plants due to its advantages such as high desulfurization efficiency, mature technology, and stable operation. During the entire desulfurization process, in order to maintain the reactivity of the slurry in the absorption tower, ensure the continuous and efficient desulfurization reaction, and inhibit the corrosion of equipment such as absorption towers, pipelines, and pumps, it is necessary to periodically discharge a certain amount of desulfurization wastewater to control the concentration of various harmful impurities in the slurry and ensure the long-term stable operation of the desulfurization system.

[0003] Desulfurization wastewater is a byproduct of the desulfurization process and has an extremely complex composition. Chloride ions are a key control indicator for wastewater treatment and system operation. Chloride ions mainly originate from coal combustion, makeup water, and process water, and they continuously accumulate in the desulfurization slurry. Increased concentration can cause severe pitting and stress corrosion to metal equipment such as absorption towers, pipelines, and pumps, shortening equipment lifespan and increasing maintenance costs. It can also reduce the reactivity of the desulfurization slurry, affecting desulfurization efficiency, and lowering the quality of gypsum byproducts, thus reducing their resource utilization value. Therefore, effective monitoring and precise control of chloride ion concentration in desulfurization wastewater is crucial for ensuring the safe, environmentally friendly, and economical operation of coal-fired power plants, directly impacting environmental emission compliance, equipment safety, and overall operational efficiency.

[0004] Currently, the monitoring and control of chloride ion concentration in desulfurization wastewater typically employs intermittent monitoring and control via manual laboratory testing, online monitoring and control using online ion chromatography, and indirect estimation control based on a single correlation parameter. None of these methods can achieve accurate, timely, and stable monitoring and control of chloride ion concentration, thus hindering the safe, economical, and environmentally friendly operation of the desulfurization system.

[0005] Specifically, the intermittent monitoring and control method of manual laboratory testing involves operators periodically taking samples on-site and sending them to the laboratory for chemical analysis to detect chloride ion concentration. Parameters such as wastewater discharge are then adjusted based on the test results. This method suffers from significant detection lag; from sampling and testing to obtaining results, it typically takes several hours. The control is based on lagging data and cannot respond to changes in operating conditions in real time. The system is often in a passive or inefficient state of control, easily leading to waste of water resources and reagents. It also carries the risk of chloride ion accumulation exceeding standards, exacerbating equipment corrosion, and exceeding environmental standards. In the online monitoring and control method using online ion chromatography, the instrument can achieve continuous online detection of chloride ion concentration, significantly improving timeliness compared to manual testing. However, it has significant application drawbacks: firstly, the equipment purchase and maintenance costs are high. The instruments are expensive, requiring regular calibration and consumable replacement, making maintenance difficult and costly, which is difficult for small and medium-sized coal-fired power plants to afford. Secondly, they have poor adaptability to operating conditions. Desulfurization wastewater has a high solids content and is highly corrosive, which can easily cause malfunctions such as blockage of sampling pipelines and contamination of detection electrodes. The reliability and long-term stability of the instruments are poor, making it difficult to use them continuously and stably in industrial sites. The indirect estimation control method based on a single correlation parameter estimates the concentration by establishing a correlation model between conductivity and chloride ion concentration. It does not require complex detection equipment, is low in cost, and is easy to operate. However, this method has low prediction accuracy. The coexistence of multiple ions in the desulfurization slurry can cause mutual interference, which can easily cause the correspondence between conductivity and chloride ion concentration to drift, resulting in a large estimation error. It can only be used for trend judgment and cannot meet the requirements for precise control of chloride ion concentration.

[0006] In summary, existing technologies for monitoring and controlling chloride ions in desulfurization wastewater all have inherent defects and cannot meet the needs of refined, intelligent, low-cost, and stable operation of desulfurization systems in coal-fired power plants. Summary of the Invention

[0007] In view of the technical problems existing in the prior art, the present invention provides a method and device for water quality prediction and optimization control of desulfurization wastewater in thermal power plants, so as to solve the problem that the existing methods of monitoring and controlling chloride ions in desulfurization wastewater are difficult to achieve accurate, timely and stable monitoring and regulation of chloride ion concentration, which restricts the safe, economical and environmentally friendly operation of desulfurization systems.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows: This invention provides a method for predicting and optimizing the water quality of desulfurization wastewater from thermal power plants, comprising: Acquire real-time data of process variables of the target desulfurization system; preprocess and feature select the real-time data of process variables of the target desulfurization system to obtain real-time data with strong correlation to chloride ion concentration; Based on real-time data related to chloride ion concentration features and combined with a pre-built chloride ion concentration prediction model, the real-time predicted value of chloride ion concentration is obtained; wherein, the pre-built chloride ion concentration prediction model is a trained CNN-LSTM hybrid neural network model. Based on the real-time predicted value of chloride ion concentration, combined with the preset control logic, the dosing equipment in the target desulfurization system is optimized and controlled.

[0009] Furthermore, real-time data on process variables of the target desulfurization system includes slurry state parameters in the absorber, flue gas parameters at the inlet of the absorber, equipment operating parameters in the target desulfurization system, and material flow parameters in the target desulfurization system.

[0010] Furthermore, the process of preprocessing and feature selection of real-time data of process variables in the target desulfurization system to obtain real-time data with strong correlation to chloride ion concentration is as follows: The real-time process variable data of the target desulfurization system are processed by filling missing values, removing outliers, smoothing data and normalizing data to obtain preprocessed real-time process variable data of the target desulfurization system. Grey relational analysis was used to screen the real-time process variable data of the pretreated target desulfurization system to obtain real-time data with strong correlation characteristics with chloride ion concentration.

[0011] Furthermore, the construction process of the pre-constructed chloride ion concentration prediction model is as follows: Acquire historical data of process variables of the target desulfurization system; preprocess and feature select the historical data of process variables of the target desulfurization system to obtain historical data with strong correlation to chloride ion concentration. A training dataset was constructed based on historical data showing a strong correlation with chloride ion concentration and pre-obtained experimental measurements of chloride ions in desulfurization wastewater. Based on the training dataset, the CNN-LSTM hybrid neural network model is trained using the Bayesian optimization algorithm to obtain a pre-built chloride ion concentration prediction model.

[0012] Furthermore, based on the real-time predicted value of chloride ion concentration, combined with the pre-constructed chloride ion concentration-dosage mapping model, the dosage of the reagent required for the target desulfurization system is obtained; Based on the required dosage of chemicals for the target desulfurization system, the dosing equipment in the target desulfurization system is optimized and controlled.

[0013] Furthermore, based on the real-time predicted value of chloride ion concentration and combined with preset control logic, the process of optimizing the control of the dosing equipment in the target desulfurization system is as follows: Based on the real-time predicted value of chloride ion concentration, combined with the pre-constructed chloride ion concentration-dosage mapping model, the required dosage of reagent for the target desulfurization system is obtained; The real-time predicted value of chloride ion concentration is compared with the pre-measured laboratory value of chloride ion concentration to obtain the chloride ion concentration deviation. Based on the chloride ion concentration deviation, and combined with the preset proportional coefficient and integral coefficient, the feedback correction value is calculated. Based on the dosage of the reagents required by the target desulfurization system and the feedback correction value, the feedback-optimized dosage of the reagents required by the target desulfurization system is obtained; Based on feedback from the target desulfurization system regarding the required dosage of chemicals, the dosing amount in the target desulfurization system is optimized, and the dosing equipment in the target desulfurization system is controlled in an optimized manner.

[0014] This invention also provides a water quality prediction and optimization control system for desulfurization wastewater from thermal power plants, used to implement the aforementioned water quality prediction and optimization control method for desulfurization wastewater from thermal power plants, comprising: The data acquisition module is used to acquire real-time data of process variables of the target desulfurization system; it preprocesses and performs feature selection on the real-time data of process variables of the target desulfurization system to obtain real-time data with strong correlation to chloride ion concentration. The chloride ion concentration prediction module is used to predict the real-time value of chloride ion concentration based on real-time data related to chloride ion concentration features and in combination with a pre-built chloride ion concentration prediction model; wherein, the pre-built chloride ion concentration prediction model is a trained CNN-LSTM hybrid neural network model. The optimized control module is used to optimize the control of the dosing equipment in the target desulfurization system based on the real-time predicted value of chloride ion concentration and in combination with preset control logic.

[0015] The present invention also provides an electronic device, comprising: A processor is used to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, performs the method for water quality prediction and optimization control of desulfurization wastewater from thermal power plants.

[0016] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method for water quality prediction and optimization control of desulfurization wastewater from thermal power plants.

[0017] The present invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the aforementioned method for water quality prediction and optimization control of desulfurization wastewater from thermal power plants.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention provides a method for predicting and optimizing the water quality of desulfurization wastewater from thermal power plants. It achieves accurate online prediction of chloride ion concentration in the wastewater and, based on the high-precision prediction, enables closed-loop optimization control of chemical dosing and discharge. This effectively reduces wastewater discharge and chemical consumption, achieving safe, economical, and environmentally friendly coordinated operation of the desulfurization system. Specifically, based on mechanistic analysis and data-driven modeling techniques, and using real-time data related to chloride ion concentration characteristics, combined with a pre-constructed chloride ion concentration prediction model, it achieves online, real-time, and high-precision prediction of chloride ion concentration in desulfurization wastewater. This provides a basis for combustion optimization and precise water system regulation. The system provides real-time data support for the control process; based on the real-time predicted value of chloride ion concentration, it optimizes the control of the dosing equipment in the desulfurization system. Combined with preset control logic, the optimized control of the dosing equipment can respond to changes in operating conditions in a timely manner, achieve precise and stable regulation of chloride ion concentration, effectively inhibit the corrosion of equipment caused by excessive chloride ion accumulation, ensure the reactivity and desulfurization efficiency of the desulfurization slurry, improve the quality of gypsum by-products, and reduce water and chemical waste. This ensures the safe, environmentally friendly, and economical operation of the desulfurization system, meeting the needs of refined, intelligent, low-cost, and stable operation of desulfurization systems in coal-fired power plants.

[0019] Furthermore, based on historical data of process variables of the target desulfurization system, a training dataset was constructed after preprocessing and feature selection. This ensured a high degree of consistency between the training data and actual operating conditions, enabling the trained model to better adapt to complex on-site conditions. The CNN-LSTM hybrid neural network model was trained using a Bayesian optimization algorithm, which quickly found the optimal parameter combination, improving training efficiency and prediction accuracy. Simultaneously, experimentally measured chloride ion values ​​were used as label data to ensure the accuracy of model training. This enabled the trained CNN-LSTM hybrid neural network model to accurately capture the complex nonlinear relationship between chloride ion concentration and related feature parameters, providing reliable model support for real-time and accurate prediction of chloride ion concentration.

[0020] Furthermore, by using a pre-constructed chloride ion concentration-dosage mapping model, precise matching between chloride ion concentration and reagent dosage is achieved. Based on real-time predicted values, the reagent dosage is dynamically determined, which can respond promptly to changes in chloride ion concentration and realize dynamic adjustment of the dosing process. This ensures that the reagent dosage can effectively control the chloride ion concentration, reduces reagent consumption, lowers operating costs, and avoids excessive chloride ion accumulation and equipment corrosion caused by insufficient reagent dosage, effectively ensuring the stable and efficient operation of the desulfurization system.

[0021] Furthermore, by comparing the real-time predicted value of chloride ion concentration with the test value, the concentration deviation is accurately obtained, providing a reliable basis for feedback correction of dosing control. This avoids minor deviations that may occur due to a single predicted value, thus improving the accuracy of dosing control. By combining the preset proportional coefficient and integral coefficient to calculate the feedback correction value, dynamic correction of the dosage of the reagent is achieved, which can effectively suppress chloride ion concentration fluctuations caused by operating condition fluctuations and avoid lag and overshoot phenomena in the dosing process. By controlling the dosing equipment through feedback optimization of the dosage, closed-loop precise regulation of chloride ion concentration is achieved, which effectively improves the stability and reliability of the desulfurization system, effectively inhibits equipment corrosion, ensures desulfurization efficiency, improves gypsum quality, and at the same time minimizes the waste of reagents and water resources.

[0022] The water quality prediction and optimization control system, electronic equipment, computer-readable storage medium, and computer program products for desulfurization wastewater from thermal power plants provided by this invention possess all the advantages of the aforementioned water quality prediction and optimization control methods for desulfurization wastewater from thermal power plants. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 A flowchart of a water quality prediction and optimization control method for desulfurization wastewater from thermal power plants provided by the present invention; Figure 2 A flowchart of the water quality prediction and optimization control method for desulfurization wastewater from thermal power plants provided in Example 1; Figure 3 This is a structural block diagram of the water quality prediction and optimization control system for desulfurization wastewater from thermal power plants provided in Example 2; Figure 4 This is a structural block diagram of the electronic device provided in Example 3. Detailed Implementation

[0025] To make the technical problems solved by the present invention, the technical solutions, and the beneficial effects clearer, the following specific embodiments provide a further detailed description of the present invention. It should be understood that the specific embodiments described herein are merely illustrative and are not intended to limit the scope of the invention.

[0026] To make the technical problems, technical solutions, and beneficial effects solved by this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0027] As attached Figure 1 As shown, this invention provides a method for predicting and optimizing the water quality of desulfurization wastewater from thermal power plants, comprising the following steps: Step 100: Obtain real-time data of process variables of the target desulfurization system; preprocess and feature select the real-time data of process variables of the target desulfurization system to obtain real-time data with strong correlation to chloride ion concentration.

[0028] Step 200: Based on real-time data related to chloride ion concentration features and combined with a pre-built chloride ion concentration prediction model, the real-time predicted value of chloride ion concentration is obtained; wherein, the pre-built chloride ion concentration prediction model is a trained CNN-LSTM hybrid neural network model.

[0029] Step 300: Based on the real-time predicted value of chloride ion concentration and combined with the preset control logic, optimize the control of the dosing equipment in the target desulfurization system.

[0030] In the above embodiments, by acquiring real-time process variable data of the target desulfurization system, and preprocessing and feature selection to filter out real-time data with strong correlation features to chloride ion concentration, the problem of correspondence drift caused by multiple ion interferences in the indirect estimation method based on a single correlation parameter is effectively avoided, thus improving data correlation and accuracy. The use of a trained CNN-LSTM hybrid neural network model for real-time chloride ion concentration prediction not only solves the drawbacks of detection lag and passive, coarse control in manual laboratory testing, achieving real-time and accurate prediction of chloride ion concentration, but also overcomes the high purchase and maintenance costs, poor adaptability to operating conditions, and insufficient reliability of online ion chromatographs. It overcomes the shortcomings of traditional methods by eliminating the need for complex and expensive equipment and high-frequency calibration and maintenance, thus reducing monitoring costs and adapting to industrial site conditions where desulfurization wastewater has high solids content and strong corrosiveness. Based on the real-time predicted value of chloride ion concentration, combined with preset control logic, the dosing equipment is optimized and controlled in a timely manner. This enables timely response to changes in operating conditions, achieving precise and stable control of chloride ion concentration, effectively inhibiting the corrosion of equipment caused by excessive chloride ion accumulation, ensuring the reactivity and desulfurization efficiency of the desulfurization slurry, improving the quality of gypsum by-products, reducing water and reagent waste, and ultimately ensuring the safe, environmentally friendly, and economical operation of the desulfurization system. This meets the needs of refined, intelligent, low-cost, and stable operation of desulfurization systems in coal-fired power plants.

[0031] The following specific embodiments further explain the water quality prediction and optimization control method for desulfurization wastewater from thermal power plants provided by the present invention: Example 1 As attached Figure 2 As shown in Example 1, this method provides a method for predicting and optimizing the water quality of desulfurization wastewater from thermal power plants, including the following steps: Step 1: Obtain real-time process variable data of the target desulfurization system. Specifically, real-time data of the process variables of the target desulfurization system is obtained by collecting data from physical sensors pre-installed on the target desulfurization system. These real-time process variable data include slurry state parameters in the absorber tower, flue gas parameters at the absorber tower inlet, equipment operating parameters in the target desulfurization system, and material flow parameters in the target desulfurization system.

[0032] The parameters of the slurry state in the absorption tower include the pH value, density, liquid level, and temperature of the slurry; the parameters of the flue gas at the inlet of the absorption tower include the SO2 concentration, O2 concentration, flue gas flow rate, and flue gas pressure; the equipment operating parameters in the target desulfurization system include the operating current of the slurry circulation pump, the operating current of the oxidation blower, the operating frequency of the wastewater discharge pump, and the real-time opening of the wastewater discharge regulating valve; and the material flow parameters in the target desulfurization system include the instantaneous flow rate on the process water makeup pipeline and the real-time liquid level of the wastewater tank.

[0033] In this embodiment 1, real-time data is collected on the slurry state parameters in the absorption tower, the flue gas parameters at the absorption tower inlet, the equipment operating parameters in the target desulfurization system, and the material flow parameters in the target desulfurization system. This data serves as real-time process variable data for the target desulfurization system. This ensures that the real-time process variable data comprehensively covers the key process parameters affecting chloride ion concentration changes in the desulfurization system, avoiding the information bias caused by collecting only one or partial parameters. It ensures that the acquired real-time data fully reflects the intrinsic relationship between the desulfurization system's operating conditions and chloride ion concentration, providing a comprehensive and reliable data foundation for subsequent data preprocessing, feature selection, and accurate chloride ion concentration prediction. This effectively improves the input quality of the pre-constructed chloride ion concentration prediction model, avoids prediction bias caused by missing data from the source, and guarantees the scientific validity and effectiveness of subsequent control logic.

[0034] Step 2: Preprocess and perform feature selection on the real-time process variable data of the target desulfurization system to obtain real-time data with strong correlation to chloride ion concentration. Specifically, the process is as follows: Step 21: Perform missing value filling, outlier removal, data smoothing, and normalization on the real-time process variable data of the target desulfurization system to obtain preprocessed real-time process variable data of the target desulfurization system. Among them, missing value filling, outlier removal, data smoothing, and normalization effectively eliminate interference data caused by equipment failure, operating condition fluctuations, etc. in the field-collected data, improve the integrity, stability, and consistency of the data, and avoid interference from abnormal data on subsequent prediction models.

[0035] Step 22: Using Grey Relation Analysis (GRA), the real-time process variable data of the pre-processed target desulfurization system are screened to obtain real-time data with strong correlation features to chloride ion concentration. Using Grey Relation Analysis to screen real-time data with strong correlation features to chloride ion concentration can accurately eliminate irrelevant or weakly correlated parameters, reduce the computational load of the prediction model, and strengthen the correlation between input features and chloride ion concentration. This effectively solves the problem of correlation drift caused by interference from multiple ions, significantly improves the accuracy of chloride ion concentration prediction, and provides high-quality feature input for subsequent real-time prediction and optimized control.

[0036] It should be noted that grey relational analysis is a multi-factor statistical analysis method. This method overcomes the shortcomings of using mathematical statistics for system analysis. It is applicable regardless of sample size or the presence or absence of patterns in the samples, and it involves less computation, avoiding discrepancies between quantitative and qualitative analysis results. The basic idea of ​​grey relational analysis is to determine the strength of the relationship based on the similarity of the geometric shapes of the sequence curves; the closer the curves are, the greater the correlation between the corresponding sequences, and vice versa. The main steps of grey relational analysis include: determining the analysis sequence, dimensionless transformation of variables, calculating the correlation coefficient, and calculating the correlation degree. The formula for calculating the correlation coefficient is shown below:

[0037] in, The correlation coefficient; The main variable normalization matrix; Normalized matrix of relevant variables; The resolution coefficient is set to 0.5.

[0038] Step 3: Establish a chloride ion concentration prediction model to obtain a pre-constructed chloride ion concentration prediction model; wherein, the pre-constructed chloride ion concentration prediction model is a trained CNN-LSTM hybrid neural network model.

[0039] The process of establishing a chloride ion concentration prediction model is as follows: Step 31: Obtain historical process variable data of the target desulfurization system from the historical database of the thermal power plant. It is worth noting that the historical process variable data of the target desulfurization system has the same data category as the real-time process variable data of the target desulfurization system obtained in Step 1 above, which will not be repeated here.

[0040] Step 32: Preprocess and feature select the historical data of process variables of the target desulfurization system to obtain historical data with strong correlation to chloride ion concentration. Similarly, the preprocessing and feature selection operations in step 32 are the same as those in step 2 above, and will not be repeated here.

[0041] Step 33: Obtain the experimental measurement value of chloride ions in the desulfurization wastewater; wherein, the time series characteristics of the experimental measurement value of chloride ions in the desulfurization wastewater are aligned with the time series characteristics of historical data that are strongly correlated with chloride ion concentration.

[0042] Step 34: Based on historical data with strong correlation features to chloride ion concentration and pre-acquired experimental chloride ion values ​​of desulfurization wastewater, construct a training dataset using the sliding window method; wherein, historical data with strong correlation features to chloride ion concentration are used as input features, and pre-acquired experimental chloride ion values ​​of desulfurization wastewater are used as target labels.

[0043] Using the sliding window method, the time window length is set to T during the construction of the training dataset; for example, in a 24-hour sampling period with a sampling interval of 15 minutes, T=96; historical data with strong correlation features with chloride ion concentration are reconstructed into three-dimensional data samples; the format of the three-dimensional data samples is [number of samples, T, N]; each three-dimensional data sample corresponds to a target label, which represents the chloride ion concentration after a preset delay after the end of the time window. The preset delay includes physical transmission and testing delays.

[0044] Step 35: Construct a CNN-LSTM hybrid neural network model; wherein, the CNN-LSTM hybrid neural network model includes an input layer, a CNN network module, an LSTM network module, a fully connected layer, and an output layer.

[0045] The input layer is used to receive three-dimensional inputs of shape (T, N).

[0046] The CNN network module is used to extract and reshape features from the three-dimensional input (T, N) to obtain a temporal feature sequence. The CNN network module includes convolutional layers and pooling layers. The convolutional layers use one-dimensional convolutional layers (Conv1D) to extract features from N variables at each time step. The convolutional kernels slide along the time axis but act on all variables, aiming to automatically learn the spatial relationships and local patterns between different variables at the same time. Preferably, multiple convolutional kernels can be set to extract different features. The pooling layers are used to reduce data dimensionality, highlight important features, and enhance the model's robustness to small time shifts. The pooling layers are placed after the convolutional layers and use one-dimensional max pooling.

[0047] The LSTM network module is used to take the reshaped temporal feature sequence extracted by the CNN network module as input, capture the long-term dependencies and dynamic evolution of data in the time dimension, and obtain high-level features; the LSTM network module consists of one or more long short-term memory network layers.

[0048] Fully connected layers are used to flatten the output of the last time step of the LSTM network module and connect it to one or more fully connected layers, progressively mapping high-level features to the final predicted value of chloride ion concentration.

[0049] The output layer is used to output the predicted value of chloride ion concentration; the output layer is a linearly activated neuron.

[0050] Step 36: Based on the training dataset, and combined with the Bayesian optimization algorithm, train the CNN-LSTM hybrid neural network model to obtain the trained CNN-LSTM hybrid neural network model as a pre-built chloride ion concentration prediction model.

[0051] Specifically, the process of training a CNN-LSTM hybrid neural network model includes: Step 361: Divide the training dataset according to a preset ratio to obtain the training set, validation set, and test set.

[0052] Step 362: Determine the hyperparameters to be optimized and define the objective function; specifically, define the key hyperparameters and their search range for the CNN-LSTM hybrid neural network model and training process; use the preset performance index of the model on the validation set as the objective function to be maximized by Bayesian optimization.

[0053] The hyperparameters to be optimized include the number of convolutional layer filters (conv_filters), the kernel size (kernel_size), the number of LSTM layer neurons (lstm_units), the learner learning rate (learning_rate), and the batch size (batch_size). The search range for the number of convolutional layer filters is [8, 64], the search range for the kernel size is [2, 5], the search range for the number of LSTM layer neurons is [32, 256], the search range for the learner learning rate is [1e-4, 1e-2], and the search range for the batch size is [32, 64, 128].

[0054] The model's preset performance metrics on the validation set include negative root mean square error (-RMSE) or negative mean absolute percentage error (-MAPE); each evaluation of the objective function represents training the model once with a specific set of hyperparameters and calculating its performance on the validation set.

[0055] Step 362: Based on the training set and validation set, perform Bayesian optimization to automatically find the optimal hyperparameter combination as the final model configuration; specifically, including: (1) Initialization: randomly select several sets of hyperparameters for initial evaluation and establish an initial Gaussian process surrogate model between the objective function and the hyperparameters; (2) Iterative optimization: repeat the following steps until the preset number of iterations is reached, such as 50 times; (3) Surrogate model update: update the Gaussian process model based on the existing (hyperparameter combination, performance score) observation points; (4) Maximize the acquisition function: use acquisition functions such as Expected Improvement (EI) to calculate the next hyperparameter combination that is most likely to bring performance improvement; (5) Evaluation: use the set of hyperparameters to configure the CNN-LSTM hybrid neural network model, train it on the training set, and evaluate it on the validation set to obtain a new performance score; (6) Data accumulation: add the new (hyperparameter combination, performance score) to the observation set; (7) Obtain the optimal configuration: after the iteration is completed, select the hyperparameter combination that makes the validation set performance optimal from all observations as the final model configuration.

[0056] Step 363: Training and Validation of the Final Model. Specifically, the optimal hyperparameter combination found using Bayesian optimization is used to retrain the CNN-LSTM hybrid neural network model on the complete training set. Then, the model performance is evaluated on a test set that has never been used for training and optimization to confirm that its prediction accuracy (such as RMSE, R²) meets the requirements for engineering applications. The model is then established, and the trained CNN-LSTM hybrid neural network model is obtained.

[0057] It should be noted that the trained CNN-LSTM hybrid neural network model file is integrated into the host computer or real-time database of the thermal power plant; a prediction service is deployed; the prediction service can realize data interface, data preprocessing, construction of real-time input, and execution of prediction; specifically, the real-time data interface is used to read the latest auxiliary variable data from DCS at a fixed frequency (e.g., every minute); data preprocessing is used to perform the same standardization processing on the real-time data as in the model building stage; constructing real-time input is used to obtain data from the past T time steps and construct a real-time input tensor of shape (1,T,N); executing prediction is used to call the model and output the real-time predicted value of chloride ion concentration.

[0058] Optionally, this embodiment 1 also provides an online adaptive correction operation for the model to establish a model performance drift monitoring and update mechanism; specifically including: (1) Triggering mechanism: when new laboratory test results are returned, the error between the model prediction value and the test value at the corresponding time is automatically calculated; (2) Judgment and action: if multiple consecutive errors exceed the preset threshold, or the system has experienced known major changes in operating conditions (such as major overhaul), the model update process is triggered; (3) Online update: the latest real-time data and test values ​​are used as new samples and added to the historical dataset. Transfer learning or incremental learning techniques are used to quickly fine-tune only the last few layers or the entire network of the model so that the model can adapt to new operating conditions and maintain long-term prediction accuracy.

[0059] Step 4: Based on real-time data related to chloride ion concentration characteristics and combined with a pre-built chloride ion concentration prediction model, the real-time predicted value of chloride ion concentration is obtained. Specifically, based on real-time data related to chloride ion concentration characteristics, a real-time input tensor of shape (1,T,N) is constructed; this (1,T,N) real-time input tensor is used as the pre-built chloride ion concentration prediction model to output the real-time predicted value of chloride ion concentration.

[0060] Step 5: Based on the real-time predicted value of chloride ion concentration and combined with the preset control logic, optimize the control of the dosing equipment in the target desulfurization system. Specifically, the steps are as follows: Step 51: Based on chemometrics and historical operational data, establish a quantitative relationship between the predicted chloride ion concentration and the required dosage, obtaining a pre-constructed chloride ion concentration-dosage mapping model; the pre-constructed chloride ion concentration-dosage mapping model is as follows:

[0061] in, The dosage of reagents required for the target desulfurization system; This is a proportionality coefficient, obtained by fitting historical data or calculating based on chemical equations; This is a real-time predicted value for chloride ion concentration; The baseline dosage is used to neutralize the inherent heavy metal background levels in the system.

[0062] Step 52: Based on the real-time predicted chloride ion concentration, and combined with the pre-built chloride ion concentration-dosage mapping model, obtain the required dosage of reagent for the target desulfurization system. It should be noted that the pre-built chloride ion concentration-dosage mapping model is used as a feedforward controller; the real-time predicted chloride ion concentration... The real-time input is fed into the pre-constructed chloride ion concentration-dosage mapping model, and its output is directly used as the core part of the feedforward setpoint of the dosing equipment. This allows the dosing amount to be adjusted in advance before the actual chloride ion concentration changes, overcoming the slow response problem caused by the lag in laboratory testing in traditional feedback control.

[0063] Step 53: Optimize the control of the dosing equipment in the target desulfurization system according to the required dosage of the reagents; specifically, adjust the required dosage of the reagents for the target desulfurization system. The process communication protocol, such as OPC, is used to send data to the PLC or frequency converter controlling the dosing equipment. Then, the dosing equipment adds the required amount of reagent according to the target desulfurization system. Adjust its stroke or speed.

[0064] In this embodiment 1, to address model prediction errors and unknown interference, a feedback correction strategy is introduced, resulting in another implementation of step 5, as detailed below: Step 501: Based on the real-time predicted value of chloride ion concentration, and combined with the pre-constructed chloride ion concentration-dosage mapping model, obtain the dosage of the reagent required for the target desulfurization system.

[0065] Step 502: Compare the real-time predicted value of chloride ion concentration with the pre-measured laboratory value of chloride ion concentration to obtain the chloride ion concentration deviation.

[0066] Step 503: Based on the chloride ion concentration deviation and in conjunction with the preset proportional and integral coefficients, calculate the feedback correction value. Specifically, using a PI (proportional-integral) controller, calculate the feedback correction value based on the chloride ion concentration deviation and in conjunction with the preset proportional and integral coefficients in the PI controller. The feedback correction value is used to correct the proportional coefficients in the pre-constructed chloride ion concentration-dosage mapping model. and benchmark injection amount .

[0067] Step 504: Based on the required dosage of the reagents for the target desulfurization system and the feedback correction value, obtain the optimized dosage of the reagents required for the target desulfurization system; the calculation process for the optimized dosage of the reagents required for the target desulfurization system is as follows:

[0068] in, Feedback optimization of the dosage of reagents required for the target desulfurization system; This is the feedback correction value.

[0069] Step 505: Optimize the dosage based on feedback from the target desulfurization system regarding the required reagents. The dosing equipment in the target desulfurization system is optimized and controlled.

[0070] Step 506: Optimize the dosage of the reagents required for the target desulfurization system based on feedback. The process communication protocol, such as OPC, is used to send data to the PLC or frequency converter controlling the dosing equipment. Then, the dosing equipment optimizes the dosage based on feedback from the target desulfurization system regarding the required reagent dosage. Adjust its stroke or speed.

[0071] Optionally, this embodiment 1 also provides a local flow closed-loop control strategy; specifically, by monitoring the actual output flow of the dosing equipment in real time, the required amount of reagent to be added to the target desulfurization system is controlled. Or the feedback optimization of the dosage of reagents required for the target desulfurization system The system performs a comparison and then implements closed-loop flow control of the dosing equipment based on the comparison results to ensure accurate execution of commands. Secondly, the system displays the predicted concentration curve, the curve showing the difference between the setpoint and actual dosing values, and the cumulative saved dosage in real time on the host computer monitoring screen, providing decision support for operators. In addition, the system sets the required dosage of chemicals for the target desulfurization system. Or the feedback optimization of the dosage of reagents required for the target desulfurization system The upper and lower limits are set to prevent over-adjustment of the actuator or waste of reagents; at the same time, the required reagent dosage for the target desulfurization system is controlled. Or the feedback optimization of the dosage of reagents required for the target desulfurization system Rate limits are applied to the rate of change of the dosing equipment to avoid frequent start-ups and shutdowns or large step changes, thus ensuring stable system operation.

[0072] Optionally, to address slow time-varying processes such as coal type changes and catalyst aging, this embodiment 1 also introduces an adaptive optimization mechanism. Specifically, according to a preset time interval, such as weekly, preset key performance indicators (KPIs) are calculated. These KPIs include reagent consumption per unit chloride ion removal. It is then determined whether the preset KPIs deviate from the optimal range. If so, a preset small-scale Bayesian optimization process is triggered, using the performance indicator as the target, to optimize the preset proportional and integral coefficients in the PI controller and the proportional coefficients in the pre-constructed chloride ion concentration-dosage mapping model. and benchmark injection amount Fine-tuning is performed to ensure that the dosing system is always kept at its optimal operating point.

[0073] The water quality prediction and optimization control method for desulfurization wastewater in thermal power plants described in Example 1 is simple, easy to operate, and can achieve online prediction of chloride ion concentration in desulfurization wastewater based on existing monitoring systems. The real-time predicted value of chloride ion concentration is used as the input for process control of the dosing equipment in the target desulfurization system. Through preset control logic, the operating parameters of the dosing pump are dynamically adjusted to achieve closed-loop optimized operation.

[0074] Example 2 As attached Figure 3 As shown in the figure, this embodiment 3 provides a water quality prediction and optimization control system for desulfurization wastewater from thermal power plants, including a data acquisition module, a chloride ion concentration prediction module, and an optimization control module.

[0075] The data acquisition module is used to acquire real-time data of process variables of the target desulfurization system; it preprocesses and selects features from the real-time data of process variables of the target desulfurization system to obtain real-time data with features strongly correlated with chloride ion concentration.

[0076] The chloride ion concentration prediction module is used to predict the real-time value of chloride ion concentration based on real-time data related to chloride ion concentration characteristics and in combination with a pre-built chloride ion concentration prediction model.

[0077] The optimized control module is used to optimize the control of the dosing equipment in the target desulfurization system based on the real-time predicted value of chloride ion concentration and in combination with preset control logic.

[0078] Optionally, this embodiment 2 also includes a modeling module; the modeling module is used to establish a chloride ion concentration prediction model and obtain a pre-constructed chloride ion concentration prediction model; wherein, the pre-constructed chloride ion concentration prediction model is a trained CNN-LSTM hybrid neural network model.

[0079] Example 3 As attached Figure 4 As shown, this embodiment 3 provides an electronic device, including: a memory for storing a computer program; a processor for executing the computer program to implement the steps of the water quality prediction and optimization control method for desulfurization wastewater from thermal power plants; or, the processor for executing the computer program to implement the functions of each module in the above-mentioned water quality prediction and optimization control system for desulfurization wastewater from thermal power plants.

[0080] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a preset function, the instruction segments describing the execution process of the computer program in the electronic device.

[0081] The electronic device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above are examples of electronic devices and do not constitute a limitation on the electronic device. It may include more components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0082] The processor can be a central processing unit, or other general-purpose processors, digital signal processors, application-specific integrated circuits, off-the-shelf programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor, or any conventional processor, etc. The processor is the control center of the electronic device, connecting various parts of the entire electronic device through various interfaces and lines.

[0083] The memory can be used to store the computer program and / or module. The processor implements various functions of the electronic device by running or executing the computer program and / or module stored in the memory and by calling the data stored in the memory.

[0084] The memory may primarily include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a given function (such as sound playback, image playback, etc.). The data storage area may store data created based on the use of the mobile phone (such as audio data, phonebook, etc.). Furthermore, the memory may include high-speed random access memory and non-volatile memory, such as hard disks, RAM, plug-in hard disks, smart memory cards, secure digital cards, flash memory cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0085] Example 4 This embodiment 4 also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method for predicting and optimizing the water quality of desulfurization wastewater from thermal power plants.

[0086] If the module / unit of the water quality prediction and optimization control system for desulfurization wastewater in thermal power plants is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.

[0087] Based on this understanding, the present invention can implement all or part of the processes in the above-mentioned method for water quality prediction and optimization control of desulfurization wastewater from thermal power plants. This can also be accomplished by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of the above-mentioned method for water quality prediction and optimization control of desulfurization wastewater from thermal power plants. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or a preset intermediate form, etc.

[0088] The computer-readable storage medium may include any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory, random access memory, electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0089] Example 5 This embodiment 5 provides a computer product, which includes a computer program stored in a computer-readable storage medium. The processor of the electronic device reads the computer program from the computer-readable storage medium and executes the computer program, so that the electronic device can execute the water quality prediction and optimization control method for desulfurization wastewater of thermal power plants described in embodiment 1, which will not be repeated here.

[0090] It should be noted that those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above methods.

[0091] The water quality prediction and optimization control method for desulfurization wastewater in thermal power plants described in this invention constructs a CNN-LSTM hybrid neural network model to achieve soft measurement of chloride ion concentration using process variables of the desulfurization system. This fundamentally avoids the high purchase and maintenance costs and the risk of clogging and failure of online ion chromatographs, while overcoming the data lag of manual testing. Specifically, by using feature selection to eliminate redundant interference, and utilizing CNN to extract spatial coupling features between variables and LSTM to capture the temporal dependence of concentration changes, the prediction accuracy and robustness under complex multi-ion coexistence conditions are significantly improved, solving the problem of large estimation errors for a single conductivity parameter. Based on the high-precision real-time prediction of chloride ion concentration, the dosing equipment in the target desulfurization system is optimized and controlled, achieving closed-loop optimization control of dosing and discharge. This transforms the desulfurization system from a passive response to an active regulation, ensuring stable chloride ion concentration compliance, reducing equipment corrosion, effectively reducing wastewater discharge and reagent consumption, and achieving safe, economical, and environmentally friendly coordinated operation of the desulfurization system.

[0092] In this invention, by establishing a mathematical relationship model between process variables and target variables that are easy to measure online, real-time and continuous estimation of key quality or component parameters is achieved. Based on data-driven or mechanism-data fusion modeling methods, the invention effectively overcomes the lag of traditional manual testing and the reliability problems of online instruments, providing real-time data support for processes such as combustion optimization and precise water system control, thus demonstrating great potential in improving efficiency and reducing consumption and emissions.

[0093] The above embodiments are merely one of the implementation methods for achieving the technical solution of the present invention. The scope of protection claimed by the present invention is not limited to this embodiment, but also includes any variations, substitutions and other implementation methods that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention.

Claims

1. A method for predicting and optimizing the water quality of desulfurization wastewater from thermal power plants, characterized in that, include: Acquire real-time data of process variables in the target desulfurization system; The real-time process variable data of the target desulfurization system are preprocessed and feature selected to obtain real-time data with strong correlation to chloride ion concentration. Based on real-time data related to chloride ion concentration features and combined with a pre-built chloride ion concentration prediction model, the real-time predicted value of chloride ion concentration is obtained; wherein, the pre-built chloride ion concentration prediction model is a trained CNN-LSTM hybrid neural network model. Based on the real-time predicted value of chloride ion concentration, combined with the preset control logic, the dosing equipment in the target desulfurization system is optimized and controlled.

2. The method for water quality prediction and optimization control of desulfurization wastewater from thermal power plants according to claim 1, characterized in that, Real-time process variable data of the target desulfurization system, including slurry state parameters in the absorber, flue gas parameters at the inlet of the absorber, equipment operating parameters in the target desulfurization system, and material flow parameters in the target desulfurization system.

3. The method for water quality prediction and optimization control of desulfurization wastewater from thermal power plants according to claim 1, characterized in that, The process of preprocessing and feature selection of real-time process variable data from the target desulfurization system to obtain real-time data with strong correlation to chloride ion concentration is as follows: The real-time process variable data of the target desulfurization system are processed by filling missing values, removing outliers, smoothing data and normalizing data to obtain preprocessed real-time process variable data of the target desulfurization system. Grey relational analysis was used to screen the real-time process variable data of the pretreated target desulfurization system to obtain real-time data with strong correlation characteristics with chloride ion concentration.

4. The method for water quality prediction and optimization control of desulfurization wastewater from thermal power plants according to claim 1, characterized in that, The construction process of the pre-built chloride ion concentration prediction model is as follows: Acquire historical data of process variables of the target desulfurization system; preprocess and feature select the historical data of process variables of the target desulfurization system to obtain historical data with strong correlation to chloride ion concentration. A training dataset was constructed based on historical data showing a strong correlation with chloride ion concentration and pre-obtained experimental measurements of chloride ions in desulfurization wastewater. Based on the training dataset, the CNN-LSTM hybrid neural network model is trained using the Bayesian optimization algorithm to obtain a pre-built chloride ion concentration prediction model.

5. The method for water quality prediction and optimization control of desulfurization wastewater from thermal power plants according to claim 1, characterized in that, The process of optimizing the control of the dosing equipment in the target desulfurization system based on the real-time predicted value of chloride ion concentration and combined with the preset control logic is as follows: Based on the real-time predicted value of chloride ion concentration, combined with the pre-constructed chloride ion concentration-dosage mapping model, the required dosage of reagent for the target desulfurization system is obtained; Based on the required dosage of chemicals for the target desulfurization system, the dosing equipment in the target desulfurization system is optimized and controlled.

6. The method for water quality prediction and optimization control of desulfurization wastewater from thermal power plants according to claim 1, characterized in that, The process of optimizing the control of the dosing equipment in the target desulfurization system based on the real-time predicted value of chloride ion concentration and combined with the preset control logic is as follows: The real-time predicted value of chloride ion concentration is compared with the pre-measured laboratory value of chloride ion concentration to obtain the chloride ion concentration deviation. Based on the chloride ion concentration deviation, and combined with the preset proportional coefficient and integral coefficient, the feedback correction value is calculated. Based on the dosage of the reagents required by the target desulfurization system and the feedback correction value, the feedback-optimized dosage of the reagents required by the target desulfurization system is obtained; Based on feedback from the target desulfurization system regarding the required dosage of chemicals, the dosing amount in the target desulfurization system is optimized, and the dosing equipment in the target desulfurization system is controlled in an optimized manner.

7. A water quality prediction and optimization control system for desulfurization wastewater from thermal power plants, characterized in that, The method for predicting and optimizing the water quality of desulfurization wastewater from thermal power plants as described in any one of claims 1-6 includes: The data acquisition module is used to acquire real-time data of process variables of the target desulfurization system; it preprocesses and performs feature selection on the real-time data of process variables of the target desulfurization system to obtain real-time data with strong correlation to chloride ion concentration. The chloride ion concentration prediction module is used to predict the real-time value of chloride ion concentration based on real-time data related to chloride ion concentration features and in combination with a pre-built chloride ion concentration prediction model; wherein, the pre-built chloride ion concentration prediction model is a trained CNN-LSTM hybrid neural network model. The optimized control module is used to optimize the control of the dosing equipment in the target desulfurization system based on the real-time predicted value of chloride ion concentration and in combination with preset control logic.

8. An electronic device, characterized in that, include: A processor is used to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, performs the water quality prediction and optimization control method for desulfurization wastewater from thermal power plants as described in any one of claims 1-6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the water quality prediction and optimization control method for desulfurization wastewater from thermal power plants as described in any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the water quality prediction and optimization control method for desulfurization wastewater from thermal power plants as described in any one of claims 1-6.

Images