Intelligent dosing method and system for power plant desulfurization wastewater treatment

Abstract

The present application belongs to the technical field of power plant desulfurization wastewater treatment, and relates to a power plant desulfurization wastewater treatment intelligent dosing method and system. The present application collects feedforward signals such as water inflow, pH value and turbidity, uses a neural network model to predict the dosing demand in advance, and outputs the reference dosing amount before the wastewater enters the treatment unit. For multiple water quality parameters such as effluent turbidity, heavy metal concentration and neutralization tank pH value, independent fuzzy controllers are set up for deviation compensation, realizing decoupling control of four reagents, i.e. lime milk, organic sulfur, coagulant and coagulant aid. Taking the effluent standard rate and dosing cost as double optimization targets, online rolling optimization is carried out through reinforcement learning, which minimizes reagent consumption under the premise of ensuring environmental protection compliance, and realizes precise execution of instructions through metering pumps and regulating valves. The present application can independently realize rapid response, precise control and continuous optimization of the dosing amount, significantly improve the effluent standard rate, reduce reagent consumption and reduce manual intervention.

Description

Technical Field

[0001] This invention belongs to the field of power plant desulfurization wastewater treatment technology, and relates to an intelligent dosing method and system for power plant desulfurization wastewater treatment. Background Technology

[0002] Limestone-gypsum wet desulfurization is the mainstream flue gas purification technology for coal-fired power plants. The desulfurization wastewater produced by this process has a complex composition, containing pollutants such as suspended solids, heavy metals, and chlorides. Currently, the "three-tank" chemical precipitation method is commonly used for desulfurization wastewater treatment, which involves sequentially adding lime slurry, organic sulfides, flocculants, and coagulants to the neutralization tank, reaction tank, and flocculation tank, respectively.

[0003] However, existing dosing methods have the following problems: Insufficient dosing precision: Traditional control relies heavily on manual experience or simple PID regulation. Faced with drastic fluctuations in the quality and quantity of desulfurization wastewater, it is difficult to achieve precise dosing, often resulting in insufficient dosing (effluent exceeding standards) or excessive dosing (chemical waste, increased sludge production); Difficulty in coordinating multi-parameter coupling: The three dosing points of neutralization, sedimentation, and flocculation affect each other, and there are complex nonlinear coupling relationships between multiple parameters such as pH, ORP, turbidity, and flow rate. Traditional independent control cannot achieve global optimization; Severe control lag: The regulation method based on effluent quality feedback has significant lag. When water quality exceeds standards, substandard wastewater has already been generated, making proactive control impossible; Low level of intelligence: Existing systems are highly dependent on the experience of operators, lack self-learning and adaptive capabilities, and are difficult to cope with complex and changing operating conditions. Summary of the Invention The purpose of this invention is to provide an intelligent dosing method and system for treating desulfurization wastewater in power plants, so as to solve the technical problems of insufficient dosing accuracy, difficulty in coordinating multi-parameter coupling, significant lag, and low level of intelligence in the dosing method for desulfurization wastewater treatment.

[0004] To achieve the above objectives, the present invention employs the following technical solution: In a first aspect, the present invention provides an intelligent dosing method for treating desulfurization wastewater from power plants, comprising the following steps: S1, real-time acquisition of the operation data of the desulfurization wastewater treatment system, the operation data including influent flow rate, influent water quality data and effluent water quality data, the effluent water quality data being multiple, and the operation data being cleaned, filtered and normalized. S2, input the influent flow rate, influent pH value and influent turbidity into the trained feedforward neural network model, and the feedforward neural network model outputs the benchmark dosage of each dosing device in the future time period; S3, compare the multiple effluent water quality data with the corresponding set target values, calculate the corresponding deviation and the rate of change of the corresponding effluent water quality data, input the deviation and the rate of change into the corresponding fuzzy controller, the fuzzy controller outputs the dosage compensation coefficient, and uses the dosage compensation coefficient to correct the benchmark dosage to obtain the corrected dosage; S4. With the optimization objectives of maximizing the effluent compliance rate and minimizing the overall dosing cost, the modified dosing amount is optimized online using a reinforcement learning algorithm to generate the final dosing control command. S5, the final dosing control command is sent to each dosing execution unit to achieve precise dosing of the agent through metering pumps and flow regulating valves.

[0005] Furthermore, the influent water quality data includes influent pH value and influent turbidity; The effluent water quality data includes effluent turbidity, effluent heavy metal concentration, and neutralization tank pH value; The reference dosage includes the reference dosage of lime slurry, the reference dosage of organic sulfur, the reference dosage of coagulant, and the reference dosage of coagulant aid; The dosage compensation coefficients include the lime slurry dosage compensation coefficient, the organic sulfur dosage compensation coefficient, the coagulant dosage compensation coefficient, and the coagulant aid dosage compensation coefficient; The lime slurry dosage is corrected based on the lime slurry dosage compensation coefficient to obtain the corrected lime slurry dosage; the organic sulfur dosage is corrected based on the organic sulfur dosage compensation coefficient to obtain the corrected organic sulfur dosage; the coagulant dosage is corrected based on the coagulant dosage compensation coefficient to obtain the corrected coagulant dosage; and the coagulant aid dosage is corrected based on the coagulant aid dosage compensation coefficient to obtain the corrected coagulant aid dosage.

[0006] Furthermore, the feedforward neural network model is a multilayer perceptron structure, where the number of nodes in the input layer corresponds to the number of influent parameters, and the number of nodes in the output layer corresponds to the dosage of each dosing device. The feedforward neural network model is trained using historical operating data, and the actual dosing effect is used as the evaluation index.

[0007] Furthermore, the design of the fuzzy controller includes: using the pH deviation of the neutralization chamber and the rate of change of the pH deviation of the neutralization chamber as input linguistic variables, and using the dosage compensation coefficient as output linguistic variables; formulating a fuzzy control rule table, and using the centroid method for defuzzification.

[0008] Furthermore, the reinforcement learning algorithm adopts a deep deterministic policy gradient algorithm, whose state space includes influent water quality data and current dosage, action space is the adjustment amount of dosage, and reward function comprehensively considers effluent compliance and chemical cost.

[0009] Furthermore, it also includes closed-loop feedback and model optimization steps, specifically as follows: collecting effluent water quality data after chemical dosing, comparing the actual treatment effect with the expected effect, and optimizing the feedforward neural network model, the control rules of the fuzzy controller, and the reinforcement learning algorithm online or offline based on the comparison results.

[0010] Secondly, this invention provides an intelligent dosing system for treating desulfurization wastewater from a power plant, comprising a neutralization tank, a settling tank, a flocculation tank, and a clarification tank connected in sequence, including: A multi-parameter online monitoring unit is used to collect real-time operating data of the desulfurization wastewater treatment system. The operating data includes influent flow rate, influent water quality data, and effluent water quality data, and the effluent water quality data includes multiple parameters. A control system, electrically connected to the multi-parameter online monitoring unit, includes: The data preprocessing module is used to clean, filter, and normalize the running data; The feedforward prediction module is used to input the influent flow rate, influent pH value, and influent turbidity into a trained feedforward neural network model, which outputs the baseline dosage for each dosing device in the future time period. The feedback compensation module is used to compare multiple effluent water quality data with corresponding set target values, calculate the corresponding deviation and the rate of change of the corresponding effluent water quality data, input the deviation and the rate of change into the corresponding fuzzy controller, the fuzzy controller outputs the dosage compensation coefficient, and uses the dosage compensation coefficient to correct the benchmark dosage to obtain the corrected dosage. The multi-objective optimization module is used to optimize the corrected dosing amount online with the objectives of maximizing the effluent compliance rate and minimizing the overall dosing cost, and generates the final dosing control command by using a reinforcement learning algorithm. The dosing execution unit, electrically connected to the control system, includes a metering pump and a flow regulating valve, and is used to accurately add the agent according to the final dosing control command.

[0011] Furthermore, the influent water quality data includes influent pH value and influent turbidity; The effluent water quality data includes effluent turbidity, effluent heavy metal concentration, and neutralization tank pH value; The reference dosage includes the reference dosage of lime slurry, the reference dosage of organic sulfur, the reference dosage of coagulant, and the reference dosage of coagulant aid; The dosage compensation coefficients include the lime slurry dosage compensation coefficient, the organic sulfur dosage compensation coefficient, the coagulant dosage compensation coefficient, and the coagulant aid dosage compensation coefficient; The lime slurry dosage is corrected based on the lime slurry dosage compensation coefficient to obtain the corrected lime slurry dosage; the organic sulfur dosage is corrected based on the organic sulfur dosage compensation coefficient to obtain the corrected organic sulfur dosage; the coagulant dosage is corrected based on the coagulant dosage compensation coefficient to obtain the corrected coagulant dosage; and the coagulant aid dosage is corrected based on the coagulant aid dosage compensation coefficient to obtain the corrected coagulant aid dosage.

[0012] Furthermore, the multi-parameter online monitoring unit includes: a first pH meter, a flow meter, and a turbidity meter installed on the inlet pipe of the neutralization tank; a second pH meter installed in the neutralization tank; and a turbidity meter and a heavy metal analyzer installed in the clarification tank.

[0013] Furthermore, the dosing unit includes: a lime slurry metering pump, a lime slurry solution tank, a coagulant metering pump, a coagulant solution tank, an organic sulfur metering pump, an organic sulfur solution tank, a coagulant aid metering pump, and a coagulant aid solution tank; The lime slurry solution tank is connected to the neutralization tank via the lime slurry metering pump; The coagulant solution tank is connected to the settling tank via the coagulant metering pump; The organic sulfur solution tank is connected to the settling tank via the organic sulfur metering pump; The coagulant solution tank is connected to the flocculation tank via the coagulant metering pump.

[0014] Compared with the prior art, the present invention has the following beneficial effects: This invention collects feedforward signals such as influent flow rate, pH value, and turbidity, and uses a neural network model to predict dosing needs in advance. It outputs a baseline dosing amount before the wastewater enters the treatment unit, fundamentally solving the lag problem of traditional feedback control's "adjustment after exceeding standards" mechanism. For multiple water quality parameters such as effluent turbidity, heavy metal concentration, and neutralization tank pH value, independent fuzzy controllers are set up for deviation compensation, achieving decoupled control of four agents: lime slurry, organic sulfur, coagulant, and coagulant aid, overcoming the control difficulties caused by multi-parameter nonlinear coupling. With effluent compliance rate and dosing cost as dual optimization objectives, online rolling optimization is performed through reinforcement learning to minimize agent consumption while ensuring environmental compliance. Precise execution of commands is achieved through metering pumps and regulating valves. This invention enables the system to autonomously achieve rapid response, precise control, and continuous optimization of dosing dosage under conditions of drastic fluctuations in influent water quality and quantity, significantly improving effluent compliance rate, reducing agent consumption, minimizing manual intervention, and realizing intelligent operation of the desulfurization wastewater treatment process.

[0015] The system of this invention includes: a multi-parameter online monitoring unit, a data preprocessing module, a feedforward prediction module, a feedback compensation module, a multi-objective optimization module, and a dosing execution unit. These modules work synergistically to achieve precise, intelligent, and adaptive dosing for desulfurization wastewater treatment. Attached Figure Description

[0016] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a system connection diagram of the present invention.

[0017] The system includes: 1. Wastewater pump; 2. First pH meter; 3. Flow meter; 4. Turbidity meter; 5. Lime slurry metering pump; 6. Lime slurry solution tank; 7. Neutralization tank; 8. Second pH meter; 9. Coagulant metering pump; 10. Coagulant solution tank; 11. Settling tank; 12. Organic sulfur metering pump; 13. Organic sulfur solution tank; 14. Flocculation tank; 15. Coagulant aid metering pump; 16. Coagulant aid solution tank; 17. Clarifying tank; 18. Turbidity meter; 19. Heavy metal analyzer; 20. Control system. Detailed Implementation

[0018] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0019] It should be noted that the terms "first," "second," etc., in the specification and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0020] The present invention will now be described in further detail with reference to the accompanying drawings: See Figure 1 This invention discloses an intelligent dosing method for treating desulfurization wastewater from power plants, comprising the following steps: S1, Data Acquisition and Preprocessing Steps: Real-time acquisition of operational data from the desulfurization wastewater treatment system, including influent flow rate, influent water quality data, and effluent water quality data (multiple effluent water quality data points). The operational data is then cleaned, filtered, and normalized. This provides a high-quality, standardized multi-parameter data foundation for subsequent control, eliminating noise and dimensional differences.

[0021] In a preferred embodiment of the present invention, the influent water quality data includes influent pH value and influent turbidity; the effluent water quality data includes effluent turbidity, effluent heavy metal concentration and neutralization tank pH value; S2, Feedforward Prediction Step: The influent flow rate, influent pH value, and influent turbidity are input into a trained feedforward neural network model. The feedforward neural network model outputs the baseline dosing amount for each dosing device in the future time period. This invention predicts dosing demand in advance based on influent parameters and outputs the baseline dosing amount, overcoming the lag of traditional feedback control.

[0022] In a preferred embodiment of the present invention, the reference dosage includes the reference dosage of lime slurry, the reference dosage of organic sulfur, the reference dosage of coagulant, and the reference dosage of coagulant aid. In a preferred embodiment of the present invention, the feedforward neural network model is a multilayer perceptron structure, wherein the number of nodes in the input layer corresponds to the number of influent parameters, and the number of nodes in the output layer corresponds to the dosage of each dosing device; the feedforward neural network model is trained using historical operating data, and the actual dosing effect is used as the evaluation index.

[0023] S3, Feedback Compensation Step: Multiple effluent water quality data points are compared with corresponding set target values. The corresponding deviation and rate of change of the effluent water quality data are calculated. The deviation and rate of change are input into a corresponding fuzzy controller. The fuzzy controller outputs a dosing compensation coefficient. The baseline dosing dosage is corrected using the dosing compensation coefficient to obtain the corrected dosing dosage. Each reagent is independently corrected based on the effluent water quality deviation to eliminate feedforward prediction errors and unmodeled disturbances.

[0024] In a preferred embodiment of the present invention, the dosage compensation coefficient includes the dosage compensation coefficient for lime slurry, the dosage compensation coefficient for organic sulfur, the dosage compensation coefficient for coagulant, and the dosage compensation coefficient for coagulant aid. In a preferred embodiment of the present invention, the reference dosage of lime slurry is corrected based on the lime slurry dosage compensation coefficient to obtain a corrected lime slurry dosage; the reference dosage of organic sulfur is corrected based on the organic sulfur dosage compensation coefficient to obtain a corrected organic sulfur dosage; the reference dosage of coagulant is corrected based on the coagulant dosage compensation coefficient to obtain a corrected coagulant dosage; and the reference dosage of coagulant aid is corrected based on the coagulant aid dosage compensation coefficient to obtain a corrected coagulant aid dosage.

[0025] In a preferred embodiment of the present invention, the pH value of the neutralization chamber is compared with the preset target pH value of the neutralization chamber, the pH deviation and its rate of change are calculated, the pH deviation and its rate of change are input into a first fuzzy controller, the first fuzzy controller outputs a lime milk dosage compensation coefficient, and the lime milk dosage compensation coefficient is used to correct the lime milk baseline dosage to obtain the corrected lime milk dosage. In a preferred embodiment of the present invention, the concentration of heavy metals in the effluent is compared with a preset target value for heavy metals, the deviation of heavy metals and its rate of change are calculated, the deviation of heavy metals and its rate of change are input into a second fuzzy controller, the second fuzzy controller outputs an organic sulfur dosage compensation coefficient, and the organic sulfur dosage compensation coefficient is used to correct the organic sulfur baseline dosage to obtain the corrected organic sulfur dosage. In a preferred embodiment of the present invention, the effluent turbidity is compared with a preset turbidity target value, the turbidity deviation and its rate of change are calculated, and the turbidity deviation and its rate of change are input into a third fuzzy controller. The third fuzzy controller outputs a coagulant dosage compensation coefficient and a coagulant aid dosage compensation coefficient. The coagulant dosage compensation coefficient and the coagulant aid dosage compensation coefficient are used to correct the coagulant baseline dosage and the coagulant aid baseline dosage, respectively, to obtain the corrected coagulant dosage and the corrected coagulant aid dosage.

[0026] In a preferred embodiment of the present invention, the design of the fuzzy controller includes: using the pH deviation of the neutralization chamber and the rate of change of the pH deviation of the neutralization chamber as input linguistic variables, and using the dosage compensation coefficient as output linguistic variables; formulating a fuzzy control rule table, and using the centroid method for defuzzification.

[0027] S4, Multi-objective optimization step: Taking the highest effluent compliance rate and the lowest overall dosing cost as optimization objectives, a reinforcement learning algorithm is used to perform online rolling optimization of the corrected dosing dosage, generating the final dosing control command. This invention seeks the global optimum between compliance rate and cost, dynamically adjusts the dosing strategy, and avoids local optima.

[0028] In a preferred embodiment of the present invention, the reinforcement learning algorithm adopts a deep deterministic policy gradient algorithm. Its state space includes influent water quality data and current dosage, the action space is the adjustment amount of dosage, and the reward function comprehensively considers the effluent compliance and the cost of the reagents.

[0029] S5, Precise Execution Step: The final dosing control command is sent to each dosing execution unit, achieving precise dosing of the agent through metering pumps and flow regulating valves. By translating the optimized command into precise actions of the metering pumps and valves, accurate execution of the dosing dosage is ensured.

[0030] In a preferred embodiment of the present invention, a closed-loop feedback and model optimization step is further included, specifically as follows: collecting effluent water quality data after chemical dosing, comparing the actual treatment effect with the expected effect, and optimizing the feedforward neural network model, the control rules of the fuzzy controller, and the reinforcement learning algorithm online or offline based on the comparison results.

[0031] This invention first utilizes a feedforward neural network to predict dosing needs in advance, overcoming control lag. Then, multiple fuzzy controllers independently correct the dosages of lime slurry, organic sulfur, coagulants, and flocculants, eliminating deviations in each channel. Finally, reinforcement learning is used to perform multi-objective rolling optimization between effluent compliance rate and dosing cost, achieving global optimum. This invention, through a combination of feedforward prediction, feedback compensation, and multi-objective optimization, enables the system to achieve precise, rapid, and adaptive intelligent dosing even under drastic fluctuations in influent water quality, significantly improving effluent compliance rate, reducing reagent consumption and operating costs, and achieving fully automated operation.

[0032] See Figure 2 Based on the above method, the present invention also discloses an intelligent dosing system for treating desulfurization wastewater in power plants, comprising a neutralization tank 7, a settling tank 11, a flocculation tank 14, and a clarifier 17 connected in sequence, including: A multi-parameter online monitoring unit is used to collect real-time operating data of the desulfurization wastewater treatment system. The operating data includes influent flow rate, influent water quality data, and effluent water quality data, and the effluent water quality data includes multiple parameters. Control system 20, electrically connected to the multi-parameter online monitoring unit, the control system includes: The data preprocessing module is used to clean, filter, and normalize the running data; The feedforward prediction module is used to input the influent flow rate, influent pH value, and influent turbidity into a trained feedforward neural network model, which outputs the baseline dosage for each dosing device in the future time period. The feedback compensation module is used to compare multiple effluent water quality data with corresponding set target values, calculate the corresponding deviation and the rate of change of the corresponding effluent water quality data, input the deviation and the rate of change into the corresponding fuzzy controller, the fuzzy controller outputs the dosage compensation coefficient, and uses the dosage compensation coefficient to correct the benchmark dosage to obtain the corrected dosage. The multi-objective optimization module is used to optimize the corrected dosing amount online with the objectives of maximizing the effluent compliance rate and minimizing the overall dosing cost, and generates the final dosing control command by using a reinforcement learning algorithm. The dosing execution unit, electrically connected to the control system 20, includes a metering pump and a flow regulating valve, and is used to accurately add the agent according to the final dosing control command.

[0033] In a preferred embodiment of the present invention, the influent water quality data includes influent pH value and influent turbidity; The effluent water quality data includes effluent turbidity, effluent heavy metal concentration, and neutralization tank pH value; The reference dosage includes the reference dosage of lime slurry, the reference dosage of organic sulfur, the reference dosage of coagulant, and the reference dosage of coagulant aid; The dosage compensation coefficients include the lime slurry dosage compensation coefficient, the organic sulfur dosage compensation coefficient, the coagulant dosage compensation coefficient, and the coagulant aid dosage compensation coefficient; The lime slurry dosage is corrected based on the lime slurry dosage compensation coefficient to obtain the corrected lime slurry dosage; the organic sulfur dosage is corrected based on the organic sulfur dosage compensation coefficient to obtain the corrected organic sulfur dosage; the coagulant dosage is corrected based on the coagulant dosage compensation coefficient to obtain the corrected coagulant dosage; and the coagulant aid dosage is corrected based on the coagulant aid dosage compensation coefficient to obtain the corrected coagulant aid dosage.

[0034] In a preferred embodiment of the present invention, the multi-parameter online monitoring unit includes: a first pH meter 2, a flow meter 3 and a turbidity meter 4 installed in the inlet pipe of the neutralization tank 7; a second pH meter 8 installed in the neutralization tank 7; and a turbidity meter 18 and a heavy metal analyzer 19 installed in the clarifier 17.

[0035] In a preferred embodiment of the present invention, the dosing execution unit includes: a lime slurry metering pump 5, a lime slurry solution tank 6, a coagulant metering pump 9, a coagulant solution tank 10, an organic sulfur metering pump 12, an organic sulfur solution tank 13, a coagulant aid metering pump 15, and a coagulant aid solution tank 16. The lime slurry solution tank 6 is connected to the neutralization tank 7 via the lime slurry metering pump 5; The coagulant solution tank 10 is connected to the settling tank 11 via the coagulant metering pump 9; The organic sulfur solution tank 13 is connected to the settling tank 11 via the organic sulfur metering pump 12; The coagulant solution tank 16 is connected to the flocculation tank 14 via the coagulant metering pump 15.

[0036] The system of this invention collects data on the entire process of water inlet, neutralization and effluent in real time through a multi-parameter online monitoring unit. The system completes three-level decision-making processes of prediction, correction and optimization in sequence through the feedforward prediction module, feedback compensation module and multi-objective optimization module in the control system. Finally, the dosing execution unit accurately executes the instructions, forming a complete intelligent closed loop of perception, decision-making and execution.

[0037] Example 2: See Figure 1 and Figure 2 This invention provides an intelligent dosing control method and system for desulfurization wastewater treatment. The system achieves precision, intelligence, and adaptability in the dosing process through multi-parameter online monitoring, feedforward-feedback composite control, and intelligent algorithm collaboration. Specific implementation methods are as follows: S1, Data Acquisition and Preprocessing: The operating data of the desulfurization wastewater treatment system is collected in real time through a multi-parameter online monitoring unit, including influent flow rate, influent pH value, influent turbidity, neutralization tank pH value, effluent turbidity, and effluent heavy metal concentration. The collected data is then cleaned, filtered, and normalized.

[0038] The first pH meter 2, flow meter 3, and turbidity meter 4 are installed in the inlet pipe of the neutralization tank 7 to monitor the influent flow rate, influent pH value, and influent turbidity of the desulfurization wastewater. The second pH meter 8 is installed in the neutralization tank 7 to monitor the pH value in the neutralization tank 7 in real time. The turbidity meter 18 and heavy metal analyzer 19 are installed in the clarifier 17 to monitor the turbidity and heavy metal concentration in the clarifier 17 in real time.

[0039] S2, Feedforward Prediction: The collected influent flow rate and influent water quality data are input into a trained feedforward neural network model. The model outputs the baseline dosing amount for each dosing device in the future time period. The training dataset of the feedforward neural network model includes influent parameters and corresponding optimal dosing amounts from historical operating data. The feedforward neural network model is a multilayer perceptron structure, where the number of input layer nodes corresponds to the number of influent parameters, and the number of output layer nodes corresponds to the dosing amount for each dosing device. The model is trained using historical operating data, and the actual dosing effect is used as the evaluation index.

[0040] S3, Feedback Compensation: The collected effluent water quality data is compared with the set target value to calculate the deviation; the deviation and its rate of change are input into the fuzzy controller, and the fuzzy controller outputs the dosage compensation coefficient; the baseline dosage is corrected using the compensation coefficient to obtain the corrected dosage. The design of the fuzzy controller includes: using the effluent pH deviation E and the deviation rate of change EC as input linguistic variables, and the dosage compensation coefficient U as the output linguistic variable; formulating a fuzzy control rule table, and using the centroid method for defuzzification.

[0041] S4, Multi-objective optimization: With the optimization objectives of maximizing the effluent compliance rate and minimizing the overall dosing cost, a reinforcement learning algorithm is used to perform online rolling optimization of the corrected dosing dosage, generating the final dosing control command. The reinforcement learning algorithm employs a deep deterministic policy gradient algorithm. The state space includes influent parameters and the current dosing dosage, the action space represents the adjustment amount of the dosing dosage, and the reward function comprehensively considers both effluent compliance and chemical cost.

[0042] S5, Precise Execution: The dosing control command generated in step S4 is sent to each dosing execution unit, and the dosing of the agent is precisely achieved through high-precision metering pumps and flow regulating valves.

[0043] S6, Closed-loop feedback and model optimization: Collect effluent water quality data after chemical dosing, compare the actual treatment effect with the expected effect, and optimize the feedforward neural network model, fuzzy control rules and reinforcement learning model online or offline based on the comparison results.

[0044] The control system 20 includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of the above-described intelligent dosing control method for desulfurization wastewater.

[0045] This invention solves the technical problems of lagging dosing control, insufficient precision, and low level of intelligence in existing dosing systems, and significantly improves the compliance rate, economy, and automation level of desulfurization wastewater treatment. It is particularly suitable for the intelligent upgrading of desulfurization wastewater treatment in coal-fired power plants.

[0046] The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of this invention.

Claims

1. A smart dosing method for treating desulfurization wastewater from power plants, characterized in that, Includes the following steps: S1, real-time acquisition of the operation data of the desulfurization wastewater treatment system, the operation data including influent flow rate, influent water quality data and effluent water quality data, the effluent water quality data being multiple, and the operation data being cleaned, filtered and normalized. S2, input the influent flow rate, influent pH value and influent turbidity into the trained feedforward neural network model, and the feedforward neural network model outputs the benchmark dosage of each dosing device in the future time period; S3, compare the multiple effluent water quality data with the corresponding set target values, calculate the corresponding deviation and the rate of change of the corresponding effluent water quality data, input the deviation and the rate of change into the corresponding fuzzy controller, the fuzzy controller outputs the dosage compensation coefficient, and uses the dosage compensation coefficient to correct the benchmark dosage to obtain the corrected dosage; S4. With the optimization objectives of maximizing the effluent compliance rate and minimizing the overall dosing cost, the modified dosing amount is optimized online using a reinforcement learning algorithm to generate the final dosing control command. S5, the final dosing control command is sent to each dosing execution unit to achieve precise dosing of the agent through metering pumps and flow regulating valves.

2. The intelligent dosing method for treating desulfurization wastewater from power plants according to claim 1, characterized in that, The influent water quality data includes influent pH value and influent turbidity; The effluent water quality data includes effluent turbidity, effluent heavy metal concentration, and neutralization tank pH value; The reference dosage includes the reference dosage of lime slurry, the reference dosage of organic sulfur, the reference dosage of coagulant, and the reference dosage of coagulant aid; The dosage compensation coefficients include the lime slurry dosage compensation coefficient, the organic sulfur dosage compensation coefficient, the coagulant dosage compensation coefficient, and the coagulant aid dosage compensation coefficient; The lime slurry dosage is corrected based on the lime slurry dosage compensation coefficient to obtain the corrected lime slurry dosage. The organic sulfur baseline dosage is corrected based on the organic sulfur dosage compensation coefficient to obtain the corrected organic sulfur dosage. The coagulant dosage is corrected based on the coagulant dosage compensation coefficient to obtain the corrected coagulant dosage. The baseline dosage of the coagulant is corrected based on the coagulant dosage compensation coefficient to obtain the corrected coagulant dosage.

3. The intelligent dosing method for treating desulfurization wastewater from power plants according to claim 1, characterized in that, The feedforward neural network model is a multilayer perceptron structure, where the number of nodes in the input layer corresponds to the number of influent parameters, and the number of nodes in the output layer corresponds to the dosage of each dosing device. The feedforward neural network model is trained using historical operating data, and the actual dosing effect is used as the evaluation index.

4. The intelligent dosing method for treating desulfurization wastewater from power plants according to claim 1, characterized in that, The design of the fuzzy controller includes: using the pH deviation of the neutralization chamber and the rate of change of the pH deviation of the neutralization chamber as input linguistic variables, and the dosage compensation coefficient as output linguistic variables; formulating a fuzzy control rule table, and using the centroid method for defuzzification.

5. The intelligent dosing method for treating desulfurization wastewater from power plants according to claim 1, characterized in that, The reinforcement learning algorithm adopts a deep deterministic policy gradient algorithm. Its state space includes influent water quality data and current dosage, the action space is the adjustment amount of dosage, and the reward function comprehensively considers the effluent compliance and the cost of the reagents.

6. The intelligent dosing method for treating desulfurization wastewater from power plants according to claim 1, characterized in that, It also includes closed-loop feedback and model optimization steps, specifically as follows: collecting effluent water quality data after chemical dosing, comparing the actual treatment effect with the expected effect, and optimizing the feedforward neural network model, the control rules of the fuzzy controller, and the reinforcement learning algorithm online or offline based on the comparison results.

7. A smart dosing system for treating desulfurization wastewater from a power plant, comprising a neutralization tank (7), a settling tank (11), a flocculation tank (14), and a clarifier (17) connected in sequence, characterized in that, include: A multi-parameter online monitoring unit is used to collect real-time operating data of the desulfurization wastewater treatment system. The operating data includes influent flow rate, influent water quality data, and effluent water quality data, and the effluent water quality data includes multiple parameters. The control system (20) is electrically connected to the multi-parameter online monitoring unit, and the control system includes: The data preprocessing module is used to clean, filter, and normalize the running data; The feedforward prediction module is used to input the influent flow rate, influent pH value, and influent turbidity into a trained feedforward neural network model, which outputs the baseline dosage for each dosing device in the future time period. The feedback compensation module is used to compare multiple effluent water quality data with corresponding set target values, calculate the corresponding deviation and the rate of change of the corresponding effluent water quality data, input the deviation and the rate of change into the corresponding fuzzy controller, the fuzzy controller outputs the dosage compensation coefficient, and uses the dosage compensation coefficient to correct the benchmark dosage to obtain the corrected dosage. The multi-objective optimization module is used to optimize the corrected dosing amount online with the objectives of maximizing the effluent compliance rate and minimizing the overall dosing cost, and generates the final dosing control command by using a reinforcement learning algorithm. The dosing execution unit, which is electrically connected to the control system (20), includes a metering pump and a flow regulating valve, and is used to accurately add the agent according to the final dosing control command.

8. The intelligent dosing system for desulfurization wastewater treatment in power plants according to claim 7, characterized in that, The influent water quality data includes influent pH value and influent turbidity; The effluent water quality data includes effluent turbidity, effluent heavy metal concentration, and neutralization tank pH value; The reference dosage includes the reference dosage of lime slurry, the reference dosage of organic sulfur, the reference dosage of coagulant, and the reference dosage of coagulant aid; The dosage compensation coefficients include the lime slurry dosage compensation coefficient, the organic sulfur dosage compensation coefficient, the coagulant dosage compensation coefficient, and the coagulant aid dosage compensation coefficient; The lime slurry dosage is corrected based on the lime slurry dosage compensation coefficient to obtain the corrected lime slurry dosage. The organic sulfur baseline dosage is corrected based on the organic sulfur dosage compensation coefficient to obtain the corrected organic sulfur dosage. The coagulant dosage is corrected based on the coagulant dosage compensation coefficient to obtain the corrected coagulant dosage. The baseline dosage of the coagulant is corrected based on the coagulant dosage compensation coefficient to obtain the corrected coagulant dosage.

9. The intelligent dosing system for desulfurization wastewater treatment in power plants according to claim 7, characterized in that, The multi-parameter online monitoring unit includes: a first pH meter (2), a flow meter (3) and a turbidity meter (4) installed in the inlet pipe of the neutralization tank (7), a second pH meter (8) installed in the neutralization tank (7), and a turbidity meter (18) and a heavy metal analyzer (19) installed in the clarifier (17).

10. The intelligent dosing system for treating desulfurization wastewater in power plants according to claim 7, characterized in that, The dosing unit includes: a lime slurry metering pump (5), a lime slurry solution tank (6), a coagulant metering pump (9), a coagulant solution tank (10), an organic sulfur metering pump (12), an organic sulfur solution tank (13), a coagulant aid metering pump (15), and a coagulant aid solution tank (16). The lime slurry solution tank (6) is connected to the neutralization tank (7) via the lime slurry metering pump (5). The coagulant solution tank (10) is connected to the settling tank (11) via the coagulant metering pump (9). The organic sulfur solution tank (13) is connected to the settling tank (11) via the organic sulfur metering pump (12). The coagulant solution tank (16) is connected to the flocculation tank (14) via the coagulant metering pump (15).

Images