Wet type flue gas heat exchange system for thermal power plant boiler

Abstract

The invention discloses a wet flue gas heat exchange system for a thermal power plant boiler, and relates to the technical field of thermal power plant energy conservation and emission reduction, the system comprises a desulfurization tower, a flue gas heat exchange device, a chimney, a sensing layer, a decision-making layer and an execution layer, the sensing layer collects smoke, environment, working medium and equipment all-parameter data and preprocesses the data into a state data queue, the decision-making layer predicts the future state of the system through a long-short-term memory network model and generates an optimal control strategy in combination with a dynamic matrix control algorithm, and the execution layer adjusts actions of components such as graded spraying and a variable-frequency heat pump. And meanwhile, the system adapts to working condition changes depending on a closed-loop self-learning mechanism, self-adaptive response to variable working conditions can be achieved, waste heat recovery and energy consumption are optimized while the white elimination effect is guaranteed, the net energy consumption of the system is reduced, and the system is suitable for thermal power plant boiler flue gas treatment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy saving and emission reduction of thermal power plants, in particular to a wet flue gas heat exchange system for boilers of thermal power plants. BACKGROUND

[0002] In thermal power plants, the wet saturated flue gas discharged by the boiler contains a large amount of sensible heat and latent heat, and effective recovery of this part of heat and elimination of the "white smoke" (wet plume) at the outlet of the chimney are the focus of current energy saving and emission reduction. A variety of flue gas waste heat recovery and white elimination schemes have appeared in the prior art. For example, the patent for invention with publication number CN106352316A provides a dry-wet mixed flue gas heat exchange system and heat exchange method for power plant boilers, which performs multi-stage heat exchange by setting dry flue gas heat exchangers, wet flue gas heat exchangers and heat pumps, and utilizes the waste heat of boiler blowdown water. Another utility model patent with publication number CN204901801U discloses an integrated heat exchange device for a wet desulfurization system, which adopts multiple rows of staggered heat pipe bundles arranged at the horizontal intersection section of the flue, and uses high-temperature flue gas to heat low-temperature flue gas to increase the exhaust gas temperature and promote diffusion. These technical solutions achieve waste heat recovery or reduce visual pollution to some extent.

[0003] However, the above-mentioned prior art still has obvious limitations. The system represented by CN106352316A has a relatively fixed operating mode and relies on preset working conditions, and it is difficult to adaptively adjust according to the dynamic changes of external environmental parameters (such as environmental temperature and humidity) and the fluctuations of boiler load. Although the heat pipe heat exchange device described in CN204901801U has a simple structure, its heat exchange efficiency and white elimination effect are largely limited by the fixed physical structure and heat transfer characteristics, and lack active control means. More importantly, the existing systems generally regard maximizing waste heat recovery and minimizing white elimination energy consumption as two relatively independent or even contradictory targets, and lack an intelligent decision-making mechanism that can globally optimize these two targets under complex and variable working conditions. This leads to the fact that in actual operation, in order to ensure the white elimination effect, part of the waste heat recovery efficiency is sacrificed, or in order to recover more heat, additional energy consumption is increased, making it difficult to achieve the optimal overall energy efficiency of the system.

[0004] Therefore, there is an urgent need in the art for a wet flue gas heat exchange system that can intelligently adjust key operating variables in real time based on the perception of operating and environmental parameters, so as to dynamically optimize to achieve the minimization of the net energy consumption of the whole system while ensuring the white elimination effect. SUMMARY

[0005] The purpose of the present application is to make up for the shortcomings of the prior art, provide a wet flue gas heat exchange system for a thermal power plant boiler, by setting a perception layer, a decision layer and an execution layer, combining the state prediction ability of the long short-term memory network model, the optimization decision ability of the dynamic matrix control algorithm, and the closed-loop feedback and self-learning mechanism, the system can perceive the system operation and environmental parameters in real time, intelligently adjust the actions of spraying, heat pump and other equipment, realize adaptive regulation under variable working conditions, dynamically optimize the waste heat recovery and white elimination energy consumption while ensuring the wet plume elimination effect, and make up for the shortcomings of the prior art that is difficult to adapt to variable working conditions and lacks global collaborative optimization ability.

[0006] To solve the above technical problems, the present application provides the following technical scheme: a wet flue gas heat exchange system for a thermal power plant boiler, comprising a desulfurization tower, a flue gas heat exchange device and a chimney connected in sequence, characterized in that the system further comprises:

[0007] A perception layer is composed of a flue gas parameter sensor group arranged at the outlet of the desulfurization tower, key nodes inside the flue gas heat exchange device and the inlet of the chimney, an environmental parameter sensor group arranged at the top of the chimney, and a working medium and equipment state sensor group for monitoring the state of the working medium and the running state of the equipment, for real-time acquisition of full parameter data of system operation;

[0008] A decision layer is composed of an embedded edge computing unit deployed in a field control cabinet and an AI optimization algorithm core running thereon, the AI optimization algorithm core receives and processes full parameter data from the perception layer, predicts the system state in the short term through the built-in prediction model, and based on the prediction result, generates an optimal control strategy set containing spraying, heat pump, pressure regulation and reheating unit action instructions, with the goal of optimizing the total running energy efficiency of the system;

[0009] An execution layer is composed of an intelligent hierarchical spraying tower, a variable frequency heat pump unit, a flue gas pressure regulation device and a turbine power generation unit, and an adjustable heat medium valve, the intelligent hierarchical spraying tower includes at least one high-temperature spraying section and one low-temperature spraying section, each section is equipped with an independent variable frequency water pump, a spraying pipeline and a nozzle, the heat exchange ratio of the evaporation side and the condensation side of the variable frequency heat pump unit is controlled by the decision layer, the flue gas pressure regulation device is arranged before the flue gas inlet of the flue gas heat exchange device, the turbine power generation unit is arranged after the flue gas outlet of the flue gas heat exchange device, and the adjustable heat medium valve is used to adjust the flow of the heat medium into the flue gas reheating unit;

[0010] The output end of the decision layer is electrically connected with each controlled component of the execution layer, so that the optimal control strategy set can be executed in real time, thereby realizing dynamic global optimization of waste heat recovery and white elimination energy consumption of the system under variable working conditions.

[0011] Further, the specific composition of the perception layer is as follows:

[0012] The flue gas parameter sensor group includes multiple measuring points distributed in the desulfurization tower outlet pipeline, the connecting flue between the high-temperature spraying section and the low-temperature spraying section, the flue behind the low-temperature spraying section, and the chimney inlet, and each measuring point is provided with at least a flue gas temperature sensor, a flue gas humidity sensor, a flue gas flow meter, and a flue gas pressure sensor;

[0013] The environmental parameter sensor group is fixedly installed at a position on the top of the chimney which is not directly disturbed by flue gas, and at least includes an atmospheric temperature sensor, an atmospheric relative humidity sensor, and a wind speed instrument;

[0014] The working medium and equipment state sensor group includes water temperature sensors, flow meters, pH meters, and turbidimeters installed on the water supply pipelines and return water pipelines of each spraying section, and an electric power sensor for monitoring the input power of the variable frequency heat pump unit, a rotation speed sensor for monitoring the rotation speed of each variable frequency water pump, and a valve position sensor for monitoring the opening degree of each regulating valve;

[0015] All the sensors are in data communication with the embedded edge computing unit of the decision layer through the field bus, and the embedded edge computing unit is internally provided with a data cleaning and filtering module to perform outlier rejection, noise filtering, and timestamp alignment processing on the received original sensor data, to form a system state data queue with a consistent time reference for the AI optimization algorithm core to call.

[0016] Further, the AI optimization algorithm core of the decision layer includes a long short-term memory network model for system state prediction and a dynamic matrix control algorithm for generating a control strategy;

[0017] The long short-term memory network model takes the historical system state data queue of the current time and a period of time in the past as input, and predicts the change trend of the flue gas temperature, flue gas flow, flue gas humidity, and environmental temperature and humidity entering the flue gas heat exchange device within a specific time window in the future;

[0018] The dynamic matrix control algorithm takes the prediction output of the long short-term memory network model as feedforward information, combines the current system state, solves a finite time domain optimization problem with the minimum system net energy consumption as the target in a set period, and outputs an optimal control instruction sequence in the future several control periods, and only the instruction of the current period is issued to the execution layer;

[0019] The system net energy consumption is defined as the difference between the total power consumption of the heat pump unit, all water pumps, and pressure control devices and the power generation of the turbine power generation unit;

[0020] The structure of the long short-term memory network model comprises an input layer, at least two LSTM hidden layers and a fully connected output layer, weight parameters of which are obtained through historical operation data training and are regularly fine-tuned online using the latest operation data.

[0021] The system dynamic model used by the dynamic matrix control algorithm is obtained through step response test.

[0022] Further, the specific structure of the intelligent grading spray tower is:

[0023] The spray tower body is made of corrosion-resistant alloy, and the inside is sequentially provided from bottom to top with a flue gas inlet section, a high-temperature spraying section, a gas-water separator, a low-temperature spraying section, a secondary gas-water separator and a flue gas outlet section;

[0024] The high-temperature spraying section is arranged with at least one spraying layer, each spraying layer is composed of a plurality of large-diameter solid cone nozzles arranged at a set interval, the water supply pipeline connected to the high-temperature spraying variable frequency water pump outlet, and the high-temperature spraying water comes from the return water heated by the condensing side of the heat pump unit, and the temperature is controlled in a first set temperature range by adjusting the working condition of the heat pump unit;

[0025] The low-temperature spraying section is arranged with at least two spraying layers, each spraying layer is composed of a plurality of small-diameter hollow cone nozzles arranged at a closer interval, the water supply pipeline connected to the low-temperature spraying variable frequency water pump outlet, and the low-temperature spraying water comes from an external cooling water source, and the temperature is controlled in a second set temperature range lower than the temperature of the high-temperature spraying water by adjusting the cooling water amount;

[0026] The gas-water separator adopts a folded plate structure for removing large-size water droplets entrained in the flue gas;

[0027] A water collecting tank is arranged at the bottom of the tower, and part of the collected spraying waste water is returned to the spraying system after filtration and purification, and part is discharged for treatment.

[0028] Further, the coupling mode of the variable frequency heat pump unit and the spraying system is:

[0029] The evaporator of the heat pump unit serves as a low-temperature heat source, and the heat absorption side working medium circuit thereof exchanges heat with the return water pipeline of the low-temperature spraying section through a plate heat exchanger to absorb the low-temperature heat captured from the flue gas by the low-temperature spraying water;

[0030] The condenser of the heat pump unit serves as a high-temperature heat source, and the heat release side working medium circuit thereof is divided into two routes, one of which is connected with the water supply pipeline of the high-temperature spraying section through another plate heat exchanger for heating the high-temperature spraying water, and the other route can be switched to an additional heat exchanger for heating boiler feed water;

[0031] The heat pump unit uses a variable frequency compressor, whose operating frequency is dynamically adjusted by the decision-making level according to the optimization objectives, thereby changing the heat transfer rate and ratio between the evaporator and the condenser.

[0032] Under the direction of the decision-making level, the heat pump unit can smoothly switch between high energy efficiency ratio mode and high heating mode to adapt to the changing demand for waste heat recovery grade and quantity under different flue gas loads and environmental conditions.

[0033] Furthermore, the cooperative working mode of the flue gas pressure regulation device and the turbine power generation unit is as follows:

[0034] The flue gas pressure regulating device is a frequency-controlled axial flow fan, which is installed on the flue inlet of the flue gas heat exchange device. It is used to increase the static pressure of the flue gas according to the instructions of the decision-making level, so that the flue gas is kept in the heat exchange device at a set pressure value higher than the atmospheric pressure. This operation aims to increase the dew point temperature of the flue gas and enhance the condensation heat exchange process of water vapor in the flue gas.

[0035] The turbine power generation unit is an axial-flow expansion turbine generator, which is installed on the flue after the flue gas outlet of the flue gas heat exchange device. It uses the pressure energy of the flue gas after heat exchange and condensation depressurization to drive the turbine to rotate and generate electricity. The generated electricity is connected to the power plant's low-voltage power grid.

[0036] The decision-making level dynamically sets the outlet pressure setpoint of the pressure control device through optimization algorithms, so as to achieve the best balance between the enhanced condensation heat transfer benefits brought by pressurization and the benefits of pressurization energy consumption and turbine power generation. The balance point is found by solving a function that includes the benefits brought by pressurization energy consumption, turbine power generation and heat exchange efficiency improvement.

[0037] Furthermore, the mathematical description of the optimization problem solved by the dynamic matrix control algorithm in a rolling manner is as follows: in, Indicates the current moment. Indicates the future starting from the current moment. The incremental sequence of control variables within each control cycle. To control the number of variables, the controlled variables include the frequency of the high-temperature spray water pump, the frequency of the low-temperature spray water pump, the frequency of the heat pump compressor, the speed of the pressure regulating device, and the opening degree of the reheat valve. Indicates in Time prediction The system output vector at any given time includes exhaust gas temperature, exhaust gas humidity, and net system energy consumption. express The system output reference value at any given time is determined by the white powder elimination requirements and energy efficiency targets. To predict the length of the time domain, To control the length of the time domain, and The weights are positive definite and are used to penalize output errors and control increment changes, respectively.

[0038] Solving this optimization problem requires satisfying a series of constraints, including magnitude and rate constraints for each control variable, as well as constraints on the safe operating range of the system output.

[0039] Furthermore, the system also includes a closed-loop feedback and self-learning mechanism, specifically implemented as follows:

[0040] The embedded edge computing unit of the decision layer stores the system's long-term operating status data, control command data, and performance index data.

[0041] The system periodically compares the actual running results with the predicted values ​​of the core AI optimization algorithm and calculates the prediction error;

[0042] When the statistical characteristics of the prediction error exceed the set threshold, the model update program is triggered to fine-tune the weight parameters of the long short-term memory network model online using the latest accumulated running data, and to re-identify the system dynamic model used by the dynamic matrix control algorithm.

[0043] The performance index data includes the waste heat recovered per unit time, water consumption, net energy consumption of the system, and visual evaluation results of whitening elimination.

[0044] This mechanism ensures that the core AI optimization algorithm can adapt to changes in unit load, fuel characteristics, and equipment performance over time, maintaining optimal operating performance throughout the system's entire lifecycle.

[0045] Compared with existing technologies, this wet flue gas heat exchange system for thermal power plant boilers has the following advantages:

[0046] I. This invention sets up a sensing layer to collect real-time data on flue gas parameters, environmental parameters, and all parameters of the working medium and equipment status. The decision layer uses a long short-term memory network model to predict the future state of the system and combines a dynamic matrix control algorithm to solve the optimization problem with the goal of minimizing the system's net energy consumption. The execution layer adjusts the actions of intelligent graded spraying, variable frequency heat pump, flue gas pressure regulation, and turbine power generation unit according to the generated optimal control strategy. This enables the system to adaptively respond to boiler load fluctuations and dynamic changes in environmental parameters. Thus, while ensuring the effect of wet plume elimination, it achieves global synergistic optimization of waste heat recovery and white plume elimination energy consumption, thereby effectively improving the overall operating energy efficiency of the system.

[0047] Second, this invention constructs a heat cascade utilization structure for a variable frequency heat pump unit and an intelligent graded spray system, achieving cascaded capture and efficient utilization of low-temperature waste heat. Combined with a turbine power generation unit that recovers flue gas pressure energy and converts it into electrical energy for the system's own use or for grid connection, a closed-loop feedback and self-learning mechanism is set up to periodically update the weights of the long short-term memory network model and the system dynamic model used in the dynamic matrix control algorithm. This enables the recycling of energy consumption and ensures that the system adapts to long-term operating condition fluctuations such as equipment performance degradation and fuel characteristic changes, thereby maintaining the system's optimal operating state in the long term, reducing the system's total life cycle operating cost, and extending the stable operating cycle of the equipment.

[0048] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0050] Figure 1 This is a flowchart of the process of the present invention;

[0051] Figure 2 This is a diagram of the three-layer intelligent control architecture of the system of the present invention;

[0052] Figure 3 This is a schematic diagram of the coupling between the core heat exchange device and the energy flow of the present invention. Detailed Implementation

[0053] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0054] Example 1

[0055] like Figure 2 and Figure 3As shown in the figure, this embodiment discloses a wet flue gas heat exchange system for a thermal power plant boiler. The system is connected to the existing desulfurization process of the thermal power plant boiler and consists of a desulfurization tower, a flue gas heat exchange device, a chimney, and a sensing layer, a decision layer, and an execution layer connected in sequence. The sensing layer collects all the system's operating parameters in real time. The decision layer predicts the future state of the system based on an AI optimization algorithm and generates the optimal control strategy. The execution layer adjusts the actions of each controlled component according to the strategy. At the same time, it adapts to changes in operating conditions through closed-loop feedback and self-learning mechanisms, and finally achieves the synergy of efficient recovery of flue gas waste heat and elimination of wet plume, reduces the net energy consumption of the system, and meets the changing operating conditions of the thermal power plant.

[0056] Overall system composition:

[0057] The wet flue gas heat exchange system for a thermal power plant boiler in this embodiment has the following core physical flow path: the flue gas discharged from the thermal power plant boiler is desulfurized in a desulfurization tower, then enters the flue gas heat exchange device for heat exchange and moisture control, and finally is discharged through the chimney in compliance with standards. To achieve intelligent control of this process, the system adds a sensing layer, a decision-making layer, and an execution layer, which form a collaborative control system—the sensing layer is the data acquisition end, the decision-making layer is the control core, and the execution layer is the action execution end. The output end of the decision-making layer is electrically connected to each controlled component of the execution layer to ensure real-time transmission and execution of control commands.

[0058] The structure and data processing of the perception layer:

[0059] The sensing layer is used to collect flue gas parameters, environmental parameters, and working fluid and equipment status parameters during system operation, providing the decision-making layer with a complete and reliable data source. Its specific structure and working method are as follows:

[0060] Flue gas parameter sensor group: This group of sensors is distributed in four key measurement areas, namely the desulfurization tower outlet pipe, the connecting flue between the high-temperature spray section and the low-temperature spray section in the intelligent graded spray tower, the flue behind the low-temperature spray section, and the chimney inlet; each measurement area is equipped with a flue gas temperature sensor, a flue gas humidity sensor, a flue gas flow meter, and a flue gas pressure sensor to synchronously acquire the temperature, humidity, flow rate, and static pressure data of the flue gas in each area, reflecting the changes in the state of the flue gas within the system.

[0061] Environmental parameter sensor group: This group of sensors is fixedly installed at the top of the chimney in a position not directly affected by the flue gas. It is equipped with an atmospheric temperature sensor, an atmospheric relative humidity sensor and an anemometer to collect ambient temperature, ambient relative humidity and ambient wind speed data, providing a basis for decision-makers to judge the risk of wet plume formation and adjust the whitening strategy.

[0062] Working fluid and equipment status sensor group: This group of sensors is divided into two categories. One category is installed in the water supply and return pipelines of each spray section of the intelligent graded spray tower, including water temperature sensors, flow meters, pH meters and turbidity meters, to monitor the temperature, flow rate, acidity, alkalinity and cleanliness of the spray water; the other category is installed in each power equipment of the execution layer, including power sensors that monitor the input power of the variable frequency heat pump unit, speed sensors that monitor the speed of each variable frequency water pump, and valve position sensors that monitor the opening of each regulating valve, to reflect the operating load and operation status of the equipment.

[0063] All sensors establish data communication with the embedded edge computing unit of the decision layer through fieldbus or wireless transmission modules. The embedded edge computing unit has a data cleaning and filtering module to process the received raw sensor data: first, outliers caused by sensor failure or interference are removed, then data noise is reduced through filtering algorithms, and finally, the data collected by different sensors is timestamped to form a system status data queue with a consistent time base for the core of AI optimization algorithms to call.

[0064] The structure and operation of the execution layer:

[0065] The execution layer is used to execute the control strategies generated by the decision-making layer. It achieves flue gas heat exchange and whitening elimination by adjusting the operating status of key equipment. Its specific structure and working method are as follows:

[0066] Intelligent graded spray tower: The spray tower body is made of corrosion-resistant alloy or lined with anti-corrosion material. Internally, from bottom to top, it consists of a flue gas inlet section, a high-temperature spray section, a primary gas-liquid separator, a low-temperature spray section, a secondary gas-liquid separator, and a flue gas outlet section. The high-temperature spray section has at least one spray layer, each composed of multiple large-diameter solid cone nozzles arranged at specific intervals. The water supply pipeline is connected to the outlet of the high-temperature spray variable frequency water pump. The spray water comes from the return water heated by the condenser side of the variable frequency heat pump unit. The temperature is controlled within the first set temperature range by adjusting the heat pump unit's operating conditions, used for preliminary heat exchange and moisture regulation of the flue gas. The low-temperature spray section has at least... The system consists of two spray layers, each composed of multiple small-diameter hollow cone nozzles arranged at closer intervals. The water supply pipeline is connected to the outlet of a low-temperature spray variable frequency water pump. The spray water comes from an external cooling water source or circulating water cooled by a cooling tower. The temperature is controlled within a second set temperature range lower than the temperature of the high-temperature spray water, which is used to further cool and dehumidify the flue gas. The primary and secondary gas-liquid separators both adopt a baffle plate or swirl plate structure to remove large water droplets entrained in the flue gas and prevent water droplets from entering subsequent pipelines. A water collection tank is provided at the bottom of the spray tower. The collected spray wastewater is purified and filtered, and part of it is returned to the spray system for reuse, while part of it is discharged from the system for further treatment.

[0067] Variable frequency heat pump unit: This unit is used to realize the transfer and cascade utilization of heat within the system. Its evaporator serves as a low-temperature heat source, and the heat absorption side working fluid circuit is connected to the return water pipeline of the low-temperature spray section through a plate heat exchanger to absorb the low-temperature heat captured by the low-temperature spray water from the flue gas. The condenser serves as a high-temperature heat source, and the heat release side working fluid circuit is divided into two paths. One path is connected to the supply water pipeline of the high-temperature spray section through another plate heat exchanger to heat the high-temperature spray water. The other path can be switched to an auxiliary heat exchanger to heat the boiler feedwater or combustion air of the thermal power plant. The unit adopts a variable frequency compressor, and the operating frequency is dynamically adjusted by the decision-making level to change the heat transfer rate and ratio between the evaporator and the condenser. It can also smoothly switch between high energy efficiency mode and high heating mode to adapt to the waste heat recovery needs under different flue gas loads and environmental conditions.

[0068] Flue gas pressure regulation device and turbine power generation unit: The flue gas pressure regulation device is a frequency-controlled axial flow fan or blower, installed before the flue gas inlet of the flue gas heat exchange device. It is used to increase the static pressure of the flue gas according to the decision-making instructions, so that the flue gas is kept at a set pressure higher than the atmospheric pressure in the heat exchange device, thereby increasing the flue gas dew point temperature and enhancing water vapor condensation heat exchange. The turbine power generation unit is an axial flow or centripetal expansion turbine generator, installed after the flue gas outlet of the flue gas heat exchange device. It uses the pressure energy of the flue gas after heat exchange and condensation to drive the turbine to rotate and generate electricity. The generated electricity is fed into the power plant's low-voltage grid or directly supplied to the frequency converter equipment in the system, reducing the system's external power consumption.

[0069] Adjustable heat medium valve: This valve is used to regulate the flow rate of heat medium entering the flue gas reheat unit, thereby controlling the flue gas reheat temperature and ensuring that the flue gas temperature at the chimney outlet meets the requirements for wet plume elimination. Its opening degree is dynamically adjusted by the decision-making level based on the deviation between the predicted value and the reference value of the exhaust gas temperature.

[0070] The core of the AI ​​optimization algorithm for decision-making:

[0071] The core of the decision-making layer consists of embedded edge computing units and AI optimization algorithms running on them. The core of this algorithm includes a Long Short-Term Memory (LSTM) network model and a Dynamic Matrix Control (DMC) algorithm, which work together to predict system state and generate control strategies, as detailed below:

[0072] The structure and operation of the LSTM model: The LSTM model is used to predict key system state changes in the near future. Its structure includes an input layer, at least two LSTM hidden layers, and a fully connected output layer. The input of the model is a queue of system state data for the current moment and a period of time in the past. This queue includes parameters such as flue gas temperature, flue gas flow rate, flue gas moisture content (from the flue gas parameter sensor group), ambient temperature, ambient humidity (from the environmental parameter sensor group), spray water temperature, and heat pump input power (from the working fluid and equipment state sensor group). The output of the model is the changing trend of flue gas temperature, flue gas flow rate, flue gas moisture content, and ambient temperature and humidity entering the flue gas heat exchanger within a specific future time window (i.e., within the prediction time domain). The weight parameters of the model are pre-trained using historical system operating data and are periodically fine-tuned online using the latest operating data to ensure that the prediction accuracy adapts to long-term operating condition fluctuations such as equipment performance degradation and fuel characteristic changes.

[0073] The optimization problem and parameter explanation of the DMC algorithm: The DMC algorithm uses the predicted output of the LSTM model as feedforward information, combines it with the current system state, and solves the finite-time domain optimization problem in a rolling manner according to a set period to generate the optimal control instruction sequence for several future control cycles. Only the instruction for the current cycle is issued to the execution layer. The mathematical expression of its optimization objective function is as follows:

[0074] The specific explanations of each parameter in the above formula are as follows: : Indicates the moment of the current control cycle, serving as a time reference point to define the time range for prediction and control. : Indicates from the current moment From now on, the future The incremental sequence of control variables within a control cycle, its mathematical form is a column vector. ; to These represent the increments of each control variable in the current period. To control the number of variables, this system =5, and the corresponding control variables are the frequency of the high-temperature spray water pump, the frequency of the low-temperature spray water pump, the frequency of the heat pump compressor, the speed of the pressure regulating device, and the opening degree of the reheat valve. : Indicates the current moment For the future time( The value range is 1 to The predicted value is obtained from the system output vector; the output vector contains three core parameters, namely flue gas temperature (flue gas temperature at the chimney inlet), flue gas humidity (flue gas moisture content at the chimney inlet), and net system energy consumption (the difference between the total system power consumption and the power generation of the turbine generator unit). : Indicates the future The system output reference value at any given time is determined by two requirements: first, the white plume elimination requirement, which determines the minimum control standards for exhaust temperature and humidity based on ambient temperature and humidity (to ensure no wet plume is generated); second, the energy efficiency target, which determines the maximum control standard for the system's net energy consumption based on the energy consumption indicators of the thermal power plant. : Indicates the prediction time domain length, that is, the total number of cycles of the future system output predicted by the DMC algorithm. Its value needs to cover the main dynamic response process of the system to ensure that the prediction can reflect the state changes of the system over a long period of time in the future. : Indicates the control time domain length, i.e., the total number of periods in the control variable increment sequence generated by the DMC algorithm, typically Only for the future Optimizing the control increments over each cycle can reduce computational complexity.

[0075] : Represents the positive definite output error weight matrix. It is a diagonal matrix, and the values ​​of its diagonal elements reflect the degree of penalty for errors of different output parameters. For example, if it is necessary to prioritize the whitening effect, the weight coefficients of the corresponding flue gas temperature and flue gas humidity errors can be increased; if it is necessary to prioritize the reduction of energy consumption, the weight coefficients of the corresponding system net energy consumption error can be increased.

[0076] : Represents the positive definite control increment weight matrix. It is also a diagonal matrix. The value of its diagonal elements reflects the degree of penalty for changes in the increment of different control variables. The larger the value, the smoother the adjustment of the corresponding control variable, which can avoid the losses caused by frequent start-ups or drastic adjustments of the equipment.

[0077] Solving this optimization problem requires satisfying two types of constraints: first, control variable constraints, including amplitude constraints (the actual values ​​of each control variable must be within the rated operating range of the equipment, such as the frequency of the heat pump compressor needing to be between 20% and 100% of the rated frequency) and rate constraints (the rate of change of each control variable must be within the allowable range of the equipment, such as the valve opening adjustment per minute not exceeding 30%); second, system output constraints (system output parameters must be within the safe operating range, such as the exhaust temperature not lower than the dew point temperature to avoid pipeline corrosion, and the system net energy consumption not exceeding the design upper limit).

[0078] The system dynamic model used in the DMC algorithm is obtained in advance through step response testing or system identification methods. This model reflects the influence of changes in control variables on the system output, providing dynamic characteristic basis for solving optimization problems.

[0079] System closed-loop feedback and self-learning mechanism:

[0080] To ensure long-term system performance, this embodiment also includes a closed-loop feedback and self-learning mechanism, implemented as follows: The embedded edge computing unit of the decision layer stores the system's long-term operating status data, control command data, and performance index data (including waste heat recovered per unit time, water consumption, system net energy consumption, and whiteout visual evaluation results); the system compares the actual operating results (such as actual flue gas temperature and actual net energy consumption) with the predicted values ​​of the LSTM model at set intervals to calculate the prediction error; when the statistical characteristics of the prediction error (such as mean square error) exceed a set threshold, the model update procedure is triggered: the weight parameters of the LSTM model are fine-tuned online using the latest accumulated operating data, and the system dynamic model used by the DMC algorithm is re-identified and updated through the system; this mechanism ensures that the core of the AI ​​optimization algorithm can adapt to changes in operating conditions such as unit load fluctuations, fuel characteristic changes, and equipment performance degradation, maintaining optimal operating performance throughout the system's entire life cycle.

[0081] In summary, this embodiment achieves intelligent control of the wet flue gas heat exchange system in a thermal power plant boiler by constructing a collaborative control system consisting of a perception layer, a decision-making layer, and an execution layer. This is achieved through the precise prediction of the LSTM model and the dynamic optimization of the DMC algorithm. Specifically, the perception layer provides comprehensive parameter acquisition and preprocessing; the execution layer achieves tiered heat utilization and energy recovery through the synergy of graded spraying, variable frequency heat pumps, pressure regulation, and turbine power generation; and the decision-making layer generates control strategies adapted to different operating conditions through rolling optimization, coupled with a closed-loop self-learning mechanism to ensure long-term operational accuracy. The entire system can dynamically reduce net energy consumption while ensuring effective wet plume elimination, effectively addressing the limitations of existing technologies in adapting to varying operating conditions and insufficient global optimization capabilities, thus meeting the actual energy conservation and emission reduction needs of thermal power plants.

[0082] Example 2

[0083] like Figure 1 As shown in Example 1, this example elaborates on the specific steps of a wet flue gas heat exchange system for a thermal power plant boiler during operation. The specific steps are as follows:

[0084] 1. Flue gas introduction and initial state monitoring:

[0085] The wet saturated flue gas generated by the boiler is treated by the desulfurization tower and then introduced into the wet flue gas heat exchange system through the flue.

[0086] The flue gas parameter sensor group in the sensing layer collects real-time data on flue gas temperature, humidity, flow rate, and pressure at the desulfurization tower outlet, between the high-temperature spray section and the low-temperature spray section, behind the low-temperature spray section, and at the chimney inlet.

[0087] 2. Environmental parameters and working fluid status acquisition:

[0088] The environmental parameter sensor array collects data on atmospheric temperature, relative humidity, and wind speed at the top of the chimney.

[0089] The working fluid and equipment status sensor group synchronously monitors the spray water temperature, flow rate, pH value, turbidity, as well as equipment operating parameters such as heat pump unit power, water pump speed, and valve opening.

[0090] 3. Data preprocessing and state queue generation:

[0091] All sensor data is transmitted to the embedded edge computing unit at the decision layer via fieldbus or wireless transmission module.

[0092] Edge computing units perform outlier removal, noise filtering, and timestamp alignment on the data to form a consistent system status data queue.

[0093] 4. System state prediction and optimization strategy generation:

[0094] The core of the AI ​​optimization algorithm, the Long Short-Term Memory (LSTM) network model, is based on a queue of historical state data to predict the future short-term trends of flue gas temperature, flow rate, moisture content, and environmental parameters.

[0095] The Dynamic Matrix Control (DMC) algorithm uses the prediction results as feedforward and combines them with the current system state to solve the optimization problem with the goal of minimizing the system's net energy consumption. It generates a sequence of control commands, including adjustments to variables such as the frequency of the high-temperature spray pump, the frequency of the low-temperature spray pump, the frequency of the heat pump compressor, the speed of the pressure regulating device, and the opening degree of the reheat valve.

[0096] 5. Control command issuance and execution layer response:

[0097] The decision-making level sends the control commands for the current cycle to each device in the execution level in real time.

[0098] The execution layer adjusts according to the instructions:

[0099] The variable frequency water pump of the intelligent graded spray tower controls the spray water volume and temperature in the high-temperature and low-temperature sections.

[0100] The compressor frequency of the variable frequency heat pump unit is adjusted to regulate the heat transfer rate and mode;

[0101] The fan speed of the flue gas pressure control device is increased to enhance condensation heat transfer by increasing the flue gas pressure.

[0102] The turbine power generation unit generates electricity using depressurized flue gas, and the electricity is either fed into the power grid or used internally.

[0103] Adjustable heat medium valve opening to regulate heat medium flow in reheat unit.

[0104] 6. Flue gas treatment and emission:

[0105] The flue gas undergoes initial heat exchange and gas-water separation through high-temperature spraying in the spray tower, followed by further cooling and dehumidification through low-temperature spraying, and finally removes water droplets through secondary gas-water separation.

[0106] The treated flue gas meets the set temperature and humidity requirements and is discharged through the chimney, thus eliminating wet plumes.

[0107] 7. Closed-loop feedback and model self-learning:

[0108] The system periodically compares actual operating data (such as flue gas temperature and net energy consumption) with AI predictions to calculate the prediction error.

[0109] If the error statistics exceed the threshold, the model update procedure is triggered: the weights of the LSTM model are fine-tuned using the latest data, and the system dynamic model of the DMC algorithm is re-identified to adapt to long-term operating conditions such as load changes and equipment performance degradation.

[0110] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A wet flue gas heat exchange system for a boiler of a thermal power plant, comprising a desulfurization tower, a flue gas heat exchange device and a chimney connected in sequence, characterized in that, The system also comprises: a perception layer composed of a flue gas parameter sensor group arranged at the outlet of the desulfurization tower, the key nodes inside the flue gas heat exchange device, and the entrance of the chimney, an environmental parameter sensor group arranged at the top of the chimney, and a working medium and equipment state sensor group for monitoring the state of the working medium and the running state of the equipment, for real-time collection of full parameter data of system operation; a decision layer composed of an embedded edge computing unit deployed in the field control cabinet and an AI optimization algorithm core running thereon, the AI optimization algorithm core receiving and processing full parameter data from the perception layer, predicting the system state in the short term through the built-in prediction model, and generating an optimal control strategy set including spray, heat pump, pressure regulation, and reheating unit action instructions based on the prediction results, with the goal of optimizing the total running energy efficiency of the system; an execution layer composed of an intelligent hierarchical spray tower, a variable frequency heat pump unit, a flue gas pressure regulation device and a turbine power generation unit, and an adjustable heat medium valve, the intelligent hierarchical spray tower including at least one high-temperature spray section and one low-temperature spray section, each section being equipped with an independent variable frequency water pump, spray pipeline and nozzle, the heat exchange ratio of the evaporation side and the condensation side of the variable frequency heat pump unit being controlled by the decision layer, the flue gas pressure regulation device being arranged before the flue gas inlet of the flue gas heat exchange device, the turbine power generation unit being arranged after the flue gas outlet of the flue gas heat exchange device, and the adjustable heat medium valve being used to adjust the flow of heat medium into the flue gas reheating unit; the output end of the decision layer is electrically connected with each controlled component of the execution layer.

2. A wet flue gas heat recovery system for a boiler of a thermal power plant according to claim 1, characterized in that, The specific composition of the perception layer is as follows: The flue gas parameter sensor group includes multiple measurement points distributed in the desulfurization tower outlet pipeline, the connecting flue between the high-temperature spray section and the low-temperature spray section, the flue behind the low-temperature spray section, and the chimney entrance, each measurement point being equipped with at least a flue gas temperature sensor, a flue gas humidity sensor, a flue gas flow meter, and a flue gas pressure sensor; The environmental parameter sensor group is fixedly installed at the top of the chimney without direct interference from flue gas, and includes at least an atmospheric temperature sensor, an atmospheric relative humidity sensor, and an anemometer; The working medium and equipment state sensor group includes water temperature sensors, flow meters, pH meters, and turbidimeters installed on the water supply pipelines and return water pipelines of each spray section, as well as an electric power sensor for monitoring the input power of the variable frequency heat pump unit, a rotational speed sensor for monitoring the rotational speed of each variable frequency water pump, and a valve position sensor for monitoring the opening degree of each regulating valve; All sensors communicate data with the embedded edge computing unit of the decision layer through the field bus, and the embedded edge computing unit is internally provided with a data cleaning and filtering module for outlier rejection, noise filtering, and timestamp alignment processing of the received raw sensor data, forming a system state data queue with consistent time reference for calling by the AI optimization algorithm core.

3. A wet flue gas heat recovery system for a boiler of a thermal power plant according to claim 1, characterized in that, The AI optimization algorithm core of the decision layer includes a long short-term memory network model for system state prediction and a dynamic matrix control algorithm for generating control strategies. The long short-term memory network model takes a historical system state data queue of a current time and a past period of time as input, and predicts a change trend of flue gas temperature, flue gas flow, and flue gas moisture content entering the flue gas heat exchange device and environmental temperature and environmental humidity in a future specific time window; The dynamic matrix control algorithm takes the predicted output of the long short-term memory network model as feedforward information, combines the current system state, sets a periodic rolling to solve a finite time domain optimization problem with the minimum system net energy consumption as an objective, and thus outputs an optimal control instruction sequence in future control cycles, and only the instruction of the current cycle is issued to the execution layer; The system net energy consumption is defined as the difference between the total power consumption of the heat pump unit, all water pumps, and the pressure regulating device and the power generation of the turbine power generation unit; The structure of the long short-term memory network model comprises an input layer, at least two LSTM hidden layers, and a fully connected output layer, the weight parameters of which are obtained through historical operation data training and are regularly fine-tuned using the latest operation data; The system dynamic model used by the dynamic matrix control algorithm is obtained through step response testing.

4. A wet flue gas heat recovery system for a boiler of a thermal power plant as claimed in claim 1, wherein, The specific structure of the intelligent hierarchical spray tower is as follows: The spray tower body is made of corrosion-resistant alloy, and the inside of the spray tower body is sequentially provided from bottom to top with a flue gas inlet section, a high-temperature spraying section, a gas-water separator, a low-temperature spraying section, a secondary gas-water separator, and a flue gas outlet section; The high-temperature spraying section is arranged with at least one spraying layer, each spraying layer is composed of a plurality of large-diameter solid cone nozzles arranged at a set interval, a water supply pipeline connected to the high-temperature spraying variable frequency water pump outlet, and the high-temperature spraying water comes from the return water heated by the heat pump unit on the condensing side, and the temperature is controlled within a first set temperature range by adjusting the working condition of the heat pump unit; The low-temperature spraying section is arranged with at least two spraying layers, each spraying layer is composed of a plurality of small-diameter hollow cone nozzles arranged at a closer interval, a water supply pipeline connected to the low-temperature spraying variable frequency water pump outlet, and the low-temperature spraying water comes from an external cooling water source, and the temperature is controlled within a second set temperature range lower than the temperature of the high-temperature spraying water by adjusting the cooling water amount; The gas-water separator adopts a folded plate structure and is used to remove large-size water droplets entrained in the flue gas; A water collecting tank is arranged at the bottom of the tower to collect the spraying wastewater, and part of the collected spraying wastewater is returned to the spraying system after filtration and purification, and part of the collected spraying wastewater is discharged for treatment.

5. A wet flue gas heat recovery system for a boiler of a thermal power plant as claimed in claim 1, wherein, The coupling mode of the variable frequency heat pump unit and the spraying system is as follows: The evaporator of the heat pump unit serves as a low-temperature heat source, and a heat exchange between a working medium circuit on the heat absorption side of the evaporator and a return water pipeline of the low-temperature spraying section is performed through a plate heat exchanger to absorb low-temperature heat captured from the flue gas by the low-temperature spraying water; The condenser of the heat pump unit serves as a high-temperature heat source, and a working medium circuit on the heat release side of the condenser is divided into two routes, one of which is connected to a water supply pipeline of the high-temperature spraying section through another plate heat exchanger to heat the high-temperature spraying water, and the other route can be switched to an additional heat exchanger to heat boiler feed water; The heat pump unit adopts a variable frequency compressor, and the running frequency of the variable frequency compressor is dynamically adjusted by the decision layer according to the optimization objective, so as to change the heat carrying rate and proportion between the evaporator and the condenser. Under the decision layer instruction, the heat pump unit can be smoothly switched between high energy efficiency ratio mode and high heating capacity mode.

6. A wet flue gas heat recovery system for a boiler of a thermal power plant according to claim 1, characterized in that, The working mode of the flue gas pressure regulating device and the turbine power generation unit is as follows: The flue gas pressure regulating device is a variable frequency controlled axial flow fan, which is installed on the flue before the flue gas inlet of the flue gas heat exchange device, used to raise the static pressure of the flue gas according to the instruction of the decision layer, so that the flue gas maintains a set pressure value higher than the normal pressure in the heat exchange device, which aims to increase the dew point temperature of the flue gas and strengthen the condensation heat exchange process of water vapor in the flue gas; The turbine power generation unit is an axial flow expansion turbine generator, which is installed on the flue after the flue gas outlet of the flue gas heat exchange device, which uses the pressure energy of the flue gas after heat exchange and condensation pressure reduction to drive the turbine to rotate and generate electricity, and the generated electricity is connected to the low voltage power grid of the power plant; The decision layer dynamically sets the outlet pressure set point of the pressure regulating device through the optimization algorithm, so that the condensation heat exchange strengthening benefit brought by pressure increase and the pressure increase energy consumption and turbine power generation benefit reach the best balance.

7. A wet flue gas heat recovery system for a boiler of a thermal power plant according to claim 3, characterized in that, The mathematical description of the optimization problem solved by the dynamic matrix control algorithm is as follows: wherein, denotes the current time, denotes the future control period from the current time, is the number of control variables, including high-temperature spray water pump frequency, low-temperature spray water pump frequency, heat pump compressor frequency, pressure regulating device rotating speed, reheating valve opening degree, denotes the system output vector at predicted at , including exhaust gas temperature, exhaust gas humidity, system net energy consumption, denotes the system output reference value at , determined by the white smoke elimination demand and the energy efficiency target, is the length of the prediction time domain, is the length of the control time domain, and are positive definite weight matrices, respectively used to punish the output error and the control increment change.

8. A wet flue gas heat recovery system for a boiler of a thermal power plant according to claim 1, characterized in that, The system also contains a closed loop feedback and self learning mechanism, which is specifically implemented as: The embedded edge computing unit of the decision layer stores state data, control instruction data and performance index data of the system during long-term operation; The system compares the actual operation results with the predicted values of the AI optimization algorithm core regularly, and calculates the prediction error; When the statistical characteristics of the prediction error exceed the set threshold, the model updating program is triggered, the latest accumulated operation data is used to online fine tune the weight parameters of the long short term memory network model, and the system dynamic model used by the dynamic matrix control algorithm is re identified; The performance index data includes the waste heat recovered per unit time, water consumption, system net energy consumption and white vision evaluation results.

Images