Soot blowing control method for low-low-temperature economizer based on machine learning

Abstract

The invention provides a low-low-temperature economizer soot blowing control method based on machine learning. The low-low-temperature economizer soot blowing control method comprises the steps that key parameters of a low-low-temperature economizer in the soot blowing process are obtained; the key parameters comprise a fluid parameter, a temperature parameter, a heat exchange parameter and an ash deposition parameter; establishing each independent model based on the key parameters; the independent model comprises a temperature change rate model, a convective heat exchange model, a radiation heat exchange model and a reference heat loss model; the independent models are associated based on the heat change rate in the economizer, and an associated model is obtained; solving the correlation model, and determining actual heat loss caused by dust deposition at the current moment; a machine learning model is established, soot blowing frequency and soot blowing time are adjusted based on actual heat loss, the heat exchange efficiency of the low-low-temperature economizer is ensured, and balanced operation is achieved; by dynamically adjusting the soot blowing frequency and duration, it is ensured that the surface of the heat exchange element is kept in a good heat exchange state, soot blowing and equipment abrasion are balanced, energy consumption and cost are reduced, and efficient, economical and environment-friendly operation is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of thermal power generation, and in particular to a soot blowing control method for a low-temperature economizer based on machine learning. Background Art

[0002] In the field of thermal power generation and industrial waste heat recovery, improving energy efficiency and reducing pollutant emissions have always been important development directions; low-temperature economizers, as an important energy-saving equipment, play a key role in this process; in the early days, the design of low-temperature economizers mainly focused on simple heat exchange functions, heating feed water by recovering flue gas waste heat, thereby improving boiler efficiency; with the continuous advancement of technology, the performance requirements for low-temperature economizers are increasing;

[0003] During the operation of low-temperature economizers, dust particles in the flue gas will inevitably deposit on the surface of the heat exchange element to form an ash layer; the thermal conductivity of the ash layer is much lower than the thermal conductivity of the heat exchange element material, which seriously hinders the transfer of heat from the flue gas to the feed water, resulting in a gradual decrease in heat exchange efficiency; according to actual operation data, in some low-temperature economizers that have been in operation for a long time and have not been effectively blown off in time, the heat exchange efficiency may drop by 10%-30%; at the same time, the accumulation of ash will also increase the resistance of flue gas flow and increase the energy consumption of the fan; for example, in a low-temperature economizer of a power plant, due to severe ash accumulation, the fan energy consumption increased by about 20% over a period of time, affecting the economy of the entire system;

[0004] However, most of the existing soot blowing control methods are based on fixed time intervals or simple parameter threshold judgments, such as soot blowing when the inlet and outlet flue gas pressure difference reaches a certain value; this method cannot accurately adapt to the actual situation of soot accumulation under different working conditions; under working conditions such as load changes and fuel quality fluctuations, the formation speed and characteristics of soot accumulation will change, but the fixed soot blowing strategy cannot be adjusted in time; for example, when the ash content in the fuel suddenly increases, the soot accumulation speed will accelerate, but the traditional soot blowing control may not be able to respond in time, resulting in a continuous decrease in heat exchange efficiency; moreover, too frequent soot blowing will cause wear on the heat exchange components and shorten the service life of the equipment; and insufficient soot blowing cannot effectively solve the soot accumulation problem; according to statistics, unreasonable soot blowing control may shorten the service life of the heat exchange components by 1-3 years, increasing the equipment maintenance cost and replacement frequency;

[0005] Therefore, there is an urgent need in the art for a low-temperature economizer soot blowing control method based on machine learning to solve the above problems. Summary of the invention

[0006] The present invention provides a low-temperature economizer sootblowing control method based on machine learning, which aims to solve the above-mentioned problems existing in the prior art. By dynamically adjusting the sootblowing frequency and duration, it is ensured that the surface of the heat exchange element maintains a good heat exchange state, balances sootblowing and equipment wear, reduces energy consumption and costs, and realizes efficient, economical and environmentally friendly operation.

[0007] The present invention provides a low-temperature economizer soot blowing control method based on machine learning, comprising:

[0008] Step 1: Obtain key parameters in the soot blowing process of the low-temperature economizer; the key parameters include fluid parameters, temperature parameters, heat exchange parameters and ash accumulation parameters;

[0009] Step 2: Establishing independent models based on key parameters; the independent models include temperature change rate model, convection heat transfer model, radiation heat transfer model, and reference heat loss model;

[0010] Step 3: Correlate the independent models based on the heat change rate in the economizer to obtain a correlation model;

[0011] Step 4, solving the correlation model to determine the actual heat loss caused by dust accumulation at the current moment;

[0012] Step five: Establish a machine learning model to adjust the soot blowing frequency and time based on the actual heat loss to ensure the heat exchange efficiency of the low-temperature economizer and achieve balanced operation.

[0013] According to the low-temperature economizer soot blowing control method based on machine learning provided by the present invention, in step 1, the key parameters include:

[0014] The fluid parameters include the fluid density in the economizer, the fluid specific heat capacity at constant pressure, the effective volume of the economizer, and the volume flow rate of the fluid, which are obtained based on the pre-installed sensor equipment;

[0015] Temperature parameters include economizer inlet fluid temperature, outlet fluid temperature, economizer wall temperature and ambient temperature, which are obtained based on pre-installed sensor equipment;

[0016] The heat transfer parameters include the heat transfer area, convective heat transfer coefficient and wall emissivity in the economizer, which are determined based on experiments;

[0017] The convective heat transfer coefficient represents the influence of the fluid flow state, physical properties, wall geometry and roughness on the convective heat transfer;

[0018] The wall emissivity represents the ability of the surface of an object to emit thermal radiation, and its value range is {0, 1};

[0019] The ash accumulation parameters include the ash thickness change rate, ash density, constant pressure specific heat capacity of the ash and ash temperature change in the economizer, which are based on pre-installed sensor equipment and experimental measurements.

[0020] According to the low-temperature economizer soot blowing control method based on machine learning provided by the present invention, in step 2, the temperature change rate model is:

[0021]

[0022] Where V is the effective volume of the economizer, ρ is the fluid density, C p is the specific heat capacity of the fluid at constant pressure, is the rate of change of temperature over time;

[0023]

[0024] Where ΔQ loss is the actual heat loss caused by ash accumulation, ΔT is the temperature difference between the inlet and outlet of the economizer, and H is the volume flow rate of the fluid;

[0025] The temperature change rate model Represents the rate of temperature change within the entire effective volume.

[0026] According to the low-temperature economizer soot blowing control method based on machine learning provided by the present invention, in step 2, the convection heat transfer model is:

[0027] hA(T fluid -T wall )

[0028] Where h is the convective heat transfer coefficient, A is the heat transfer area in the economizer, T fluid is the economizer inlet fluid temperature, T wall is the economizer wall temperature;

[0029] The convection heat transfer model hA(T fluid -T wall ) represents the heat transfer process caused by the macroscopic movement of the fluid.

[0030] According to the low-temperature economizer soot blowing control method based on machine learning provided by the present invention, in step 2, the radiation heat exchange model is:

[0031]

[0032] Where, ∈ is the wall emissivity, σ is the Stefan-Boltzmann constant, A is the heat transfer area in the economizer, T wall is the economizer wall temperature, T sur is the absolute temperature of the surrounding environment;

[0033] The radiation heat transfer model Represents the heat exchange process caused by emitting and absorbing radiation energy.

[0034] According to the low-temperature economizer soot blowing control method based on machine learning provided by the present invention, in step 2, the reference heat loss model is:

[0035]

[0036] in, is the dust accumulation mass flow rate, C p , ash is the constant pressure specific heat capacity of the ash, ΔT ash is the change of dust accumulation temperature;

[0037]

[0038] Where E is the ash thickness change rate, A is the heat exchange area in the economizer, ρ ash is the dust accumulation density;

[0039] The reference heat loss model An idealized process representing the change in heat transfer due to the presence of dust deposits.

[0040] According to the low-temperature economizer soot blowing control method based on machine learning provided by the present invention, in step three, the association model is:

[0041]

[0042] The model is combined based on the heat change rate ΔQ in the economizer; That is, generate a correlation model;

[0043] Substitute the key parameters into the correlation model and solve the actual heat loss ΔQ caused by dust accumulation at the current moment. loss .

[0044] According to the low-temperature economizer soot blowing control method based on machine learning provided by the present invention, in step five, the machine learning model is:

[0045]

[0046] Among them, f new is the adjusted soot blowing frequency, f0 is the current soot blowing frequency, ΔQ loss is the actual heat loss caused by dust accumulation at the current moment; ΔQ th is the preset reference heat loss threshold, and when ΔQ loss Exceed ΔQth When , it means that the soot blowing frequency needs to be increased; k1 and k2 are both frequency adjustment coefficients, k1 is used to reflect the sensitivity of the situation when the severity of dust accumulation exceeds the threshold, and k2 is used to reflect the degree of reducing the soot blowing frequency when the dust accumulation is relatively light;

[0047] t new is the adjusted soot blowing time, t0 is the soot blowing time at the current moment, t max is the maximum soot blowing time; k3 is the time adjustment coefficient, which is used to reflect the quantitative degree of the relationship between the heat loss caused by soot accumulation and the soot blowing time;

[0048] And when t new <t min When t new =t min ;t min It is the minimum soot blowing time.

[0049] According to the low-temperature economizer soot blowing control method based on machine learning provided by the present invention, the training process of the frequency adjustment coefficient and the time adjustment coefficient includes:

[0050] Data collection: Collect several sets of historical operating data of low-temperature economizers, including actual heat loss caused by ash accumulation under different operating conditions, soot blowing frequency, soot blowing duration, and corresponding equipment operating effects; the equipment operating effects include changes in heat exchange efficiency and equipment wear;

[0051] Establish an objective function; the goal is to find a set of coefficients k1, k2, k3, so that after adjusting the soot blowing frequency and duration, the equipment can minimize equipment wear while ensuring heat exchange efficiency, and the frequency and duration of soot blowing operations are relatively reasonable;

[0052] Train to get the ideal coefficients; use the gradient descent method to calculate the partial derivatives of the objective function with respect to the coefficients k1, k2, and k3, and then update the parameters in the opposite direction of the gradient; through continuous iterations, until the objective function converges or reaches the preset number of iterations, the final trained coefficients k1, k2, and k3 are output.

[0053] According to the low-temperature economizer soot blowing control method based on machine learning provided by the present invention, the method further comprises: new and the adjusted sootblowing time t new Corresponding control of the action of the sootblowing equipment.

[0054] Compared with the prior art, the beneficial effects of this application are:

[0055] This application uses comprehensive data collection and physical property-based mathematical model analysis to accurately determine the actual heat loss caused by ash accumulation; uses machine learning models to dynamically adjust the frequency and duration of soot blowing to ensure that the surface of the heat exchange element maintains a good heat exchange state, minimizes the hindrance of ash accumulation to the heat exchange process, thereby improving the heat exchange efficiency, enabling the low-temperature economizer to continuously and efficiently recover flue gas waste heat, and improve the energy efficiency of the entire system;

[0056] This application establishes a multi-dimensional correlation model to deeply analyze the relationship between soot blowing parameters and equipment wear. On the premise of ensuring effective removal of accumulated ash to maintain heat exchange efficiency, the soot blowing frequency and duration are optimized and adjusted through a machine learning model to avoid unnecessary damage to the equipment caused by excessive soot blowing, thereby extending the service life of the low-temperature economizer and reducing equipment update and maintenance costs;

[0057] Improving heat exchange efficiency can directly reduce fuel consumption, because more flue gas waste heat is effectively recovered for heating feed water, etc. At the same time, optimizing soot blowing control avoids the increase of flue gas resistance caused by soot accumulation, thereby reducing fan energy consumption; in addition, by extending the service life of equipment and reducing the number of maintenance times, the equipment maintenance cost is reduced; these measures combined improve the economy of low-temperature economizers throughout the entire operating cycle, reduce the operating costs of enterprises, and improve economic benefits;

[0058] The efficient heat exchange process achieved in this application helps to reduce the boiler exhaust temperature and inhibit the generation of pollutants such as nitrogen oxides (NOx); at the same time, reducing fuel consumption indirectly reduces the emission of greenhouse gases such as carbon dioxide.

[0059] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0060] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0062] Figure 1 It is a flow chart of a low-temperature economizer soot blowing control method based on machine learning provided by an embodiment of the present invention; DETAILED DESCRIPTION

[0063] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0064] Embodiment 1:

[0065] The embodiment of the present invention provides a low-temperature economizer soot blowing control method based on machine learning, please refer to Figure 1 ,include:

[0066] Step 1: Obtain key parameters in the soot blowing process of the low-temperature economizer; the key parameters include fluid parameters, temperature parameters, heat exchange parameters and ash accumulation parameters;

[0067] Step 2: Establish independent models based on key parameters; the independent models include temperature change rate model, convection heat transfer model, radiation heat transfer model, and reference heat loss model;

[0068] Step 3: Correlate the independent models based on the heat change rate in the economizer to obtain a correlation model;

[0069] Step 4: Solve the correlation model to determine the actual heat loss caused by dust accumulation at the current moment;

[0070] Step five: Establish a machine learning model to adjust the soot blowing frequency and time based on the actual heat loss to ensure the heat exchange efficiency of the low-temperature economizer and achieve balanced operation.

[0071] The principles and beneficial effects of this embodiment are as follows: first, a series of key parameters are collected from the operation process of the low-temperature economizer, which are crucial for evaluating the equipment status and performance, including but not limited to fluid characteristics (such as flow rate), temperature distribution, heat exchange efficiency and ash accumulation degree; then, based on the collected data, multiple independent mathematical models are constructed, each model focuses on describing a specific aspect or process, such as the temperature change pattern over time, convection and radiation heat transfer mechanisms, and heat loss caused by ash accumulation; by analyzing the heat changes inside the economizer, the above independent models are linked to form a comprehensive correlation model; this model can more accurately reflect the overall behavioral characteristics of the system; the correlation model is solved using mathematical methods to calculate the actual heat loss caused by ash accumulation under current operating conditions; finally, a machine learning algorithm is used to dynamically adjust the soot blowing strategy (such as adjusting the frequency and duration of soot blowing) according to the actual heat loss situation, aiming to maintain optimal heat exchange efficiency while reducing unnecessary energy consumption.

[0072] In order to further optimize the above embodiment, in step 1, the key parameters include:

[0073] The fluid parameters include the fluid density in the economizer, the fluid specific heat capacity at constant pressure, the effective volume of the economizer, and the volume flow rate of the fluid, which are obtained based on the pre-installed sensor equipment;

[0074] Temperature parameters include economizer inlet fluid temperature, outlet fluid temperature, economizer wall temperature and ambient temperature, which are obtained based on pre-installed sensor equipment;

[0075] The heat transfer parameters include the heat transfer area, convective heat transfer coefficient and wall emissivity in the economizer, which are determined based on experiments;

[0076] The convective heat transfer coefficient represents the influence of the fluid flow state, physical properties, wall geometry and roughness on the convective heat transfer;

[0077] The wall emissivity represents the ability of the surface of an object to emit thermal radiation, and its value range is {0, 1};

[0078] The ash accumulation parameters include the ash thickness change rate, ash density, constant pressure specific heat capacity of the ash and ash temperature change in the economizer, which are based on pre-installed sensor equipment and experimental measurements.

[0079] It should be noted that for fluid density and specific heat capacity at constant pressure, high-precision density sensors and specific heat capacity sensors can be used to directly measure the density and specific heat capacity of the fluid in the economizer. These sensors work based on different physical principles. For example, for liquid fluids, density sensors can use the vibration principle to determine the fluid density by measuring the change in vibration frequency; specific heat capacity sensors can calculate the specific heat capacity by measuring the rate at which the fluid temperature rises under a certain heating power. During installation, the sensor is installed at the inlet pipe of the economizer to ensure that the fluid parameters entering the economizer are measured, and to ensure that the sensor is in full contact with the fluid to obtain accurate measurement values. The measured data needs to be processed and calibrated in real time. Since the properties of the fluid may be affected by factors such as temperature and pressure, the measured data needs to be corrected according to the actual working conditions. For example, in the case of water as a fluid, the measured data can be calibrated according to the temperature-density and temperature-specific heat capacity relationship tables of water to ensure the accuracy of the data.

[0080] For the effective volume of the economizer and the fluid volume flow rate, the effective volume of the economizer can be determined by geometric measurement during the equipment design and installation stage. For economizer parts with regular shapes (such as cylindrical pipe parts), the volume can be calculated based on their diameter and length; for irregular shaped parts, the segmentation method can be used to divide them into multiple parts with approximately regular shapes, and the volumes are calculated separately and then summed. In actual operation, the measurement value of the effective volume can also be verified and calibrated by injecting a known volume of liquid (such as water) and measuring the liquid level change.

[0081] Commonly used volume flow measurement methods include electromagnetic flowmeter, turbine flowmeter and ultrasonic flowmeter. Electromagnetic flowmeter is suitable for conductive liquids. Its principle is based on Faraday's law of electromagnetic induction. The flow velocity is calculated by measuring the induced electromotive force, and then the volume flow rate is obtained. The turbine flowmeter uses the fluid to drive the turbine to rotate, and the flow velocity and volume flow rate are determined by measuring the speed of the turbine. The ultrasonic flowmeter calculates the flow rate by measuring the time difference between the propagation of ultrasonic waves in the downstream and upstream of the fluid. These flowmeters should be installed on the inlet and outlet pipes of the economizer. The appropriate installation location and flowmeter type should be selected according to the pipe size and fluid characteristics to ensure the accuracy of the measurement. At the same time, the flowmeter should be calibrated and maintained regularly. For example, the electromagnetic flowmeter needs to check the cleanliness of the electrode and the magnetic field strength, and the turbine flowmeter needs to check whether the rotation of the turbine is flexible.

[0082] For the inlet and outlet fluid temperatures of the economizer, thermocouples are a commonly used temperature measurement sensor that works based on the thermoelectric effect. Insert the measuring end of the thermocouple sensor into the inlet and outlet pipes of the economizer to ensure full contact with the fluid. Thermocouples of different materials are suitable for different temperature ranges. For example, K-type thermocouples are suitable for temperature measurements from -200°C to 1350°C, and have good applicability for the common fluid temperature range in economizers. In order to improve the measurement accuracy, multiple thermocouples can be used for measurement and the average value can be taken. At the same time, the insulation and protection of the thermocouples should be done well to prevent short circuits and corrosion that affect the measurement accuracy.

[0083] Thermal resistance sensors use the property that the resistance of metal materials changes with temperature to measure temperature. Platinum resistance thermometers (Pt100 or Pt1000) are commonly used thermal resistance sensors with high measurement accuracy and good stability, especially suitable for temperature measurement in the low to medium temperature range. During installation, the thermal resistance sensor should be installed in a suitable position in the pipeline to ensure good heat exchange with the fluid, and temperature compensation should be performed to eliminate the influence of ambient temperature on the measurement results.

[0084] For the economizer wall temperature, a small thermocouple or thermal resistance sensor can be directly installed on the economizer wall through a special pasting or welding process to measure the wall temperature. This method can directly obtain the temperature information of the wall, but attention should be paid to the installation process of the sensor to ensure good thermal contact between the sensor and the wall, and will not affect the heat exchange performance and structural strength of the wall. After installation, the sensor should be encapsulated and protected to prevent erosion by smoke, dust, etc.

[0085] Infrared thermometers measure temperature by receiving infrared rays emitted from the surface of an object without direct contact with the object being measured. For the temperature measurement of the economizer wall, the infrared thermometer can be installed in a suitable position so that it can aim at the wall for temperature measurement. However, the measurement accuracy of this method may be affected by environmental factors (such as smoke obstruction, dust influence, etc.), so it needs to be calibrated and compensated according to the actual situation. At the same time, it is necessary to select a suitable infrared thermometer model to ensure that its measurement range and accuracy meet the requirements of economizer wall temperature measurement.

[0086] For the ambient temperature, use a common temperature sensor (such as a thermocouple or thermal resistor) installed in the environment around the economizer to measure the ambient temperature. The sensor should be installed in a well-ventilated location away from heat sources to ensure that the actual ambient temperature is measured. Multiple sensors can be used for distributed measurement, and then the average value is taken as the final ambient temperature value. At the same time, the sensor should be regularly checked and calibrated to prevent measurement errors caused by sensor failure or drift;

[0087] For the heat exchange area in the economizer, during the design stage of the economizer, the heat exchange area is determined by theoretical calculation based on its geometric shape (such as the diameter and length of the pipe, the size and number of fins, etc.). For economizers with simple structures (such as light tube economizers), the heat exchange area can be determined by calculating the outer surface area of ​​the pipe; for economizers with extended surfaces such as fins, it is necessary to calculate the surface area of ​​the fins and the area of ​​the connection between the fins and the pipes. After the equipment is installed, the heat exchange area can be verified by actual measurement methods, for example, using tools such as laser rangefinders to measure the size of the pipes and fins, and then recalculating the heat exchange area based on the measurement results to ensure that the design calculation value is consistent with the actual value.

[0088] For the convective heat transfer coefficient, a special experimental device can be built in the laboratory to measure the convective heat transfer coefficient. The experimental device includes a small heat exchanger that simulates the heat exchange process of the economizer, a temperature control system, a flow measurement system, etc. During the experiment, by controlling the flow rate, inlet temperature and other parameters of the fluid, measuring the temperature changes of the heat exchanger wall and the fluid, the convective heat transfer coefficient is calculated according to Newton's law of cooling. In order to improve the measurement accuracy, it is necessary to conduct multiple experiments, change different operating parameters (such as flow rate, temperature difference, etc.), and then perform regression analysis on the measurement results to obtain the convective heat transfer coefficient correlation formula applicable to different operating conditions.

[0089] For the wall emissivity, using a radiometer to measure it is a common method. Aim the radiometer at the economizer wall to measure the intensity of thermal radiation emitted by the wall. At the same time, use temperature sensors such as thermocouples to measure the wall temperature, and calculate the wall emissivity according to the Stefan-Boltzmann law. During the measurement process, it is necessary to ensure that the measurement angle and distance of the radiometer are appropriate, and the radiation interference of the surrounding environment must be shielded and corrected. In addition, the comparison method can also be used to compare the wall to be measured with the standard surface of known emissivity under the same thermal environment, and by measuring the radiation heat exchange between the two, the wall emissivity is calculated using the theoretical formula of thermal radiation exchange. However, this method requires precise control of experimental conditions to ensure that the emissivity of the standard surface is accurately known, and to consider the influence of surface conditions (such as roughness, degree of oxidation, etc.) on the emissivity.

[0090] For the rate of change of ash thickness, sensors based on optical principles, such as laser displacement sensors or optical triangulation sensors, are used. The sensors are installed at key positions of the economizer so that they can emit light beams and receive light signals reflected from the surface of ash deposits. By measuring the propagation time of the light signal or the change in the reflection angle, the distance from the sensor to the ash surface is calculated, and then the ash thickness is obtained. The rate of change of ash thickness is obtained by measuring the ash thickness regularly (such as every few minutes) and calculating the ratio of the thickness difference between adjacent measurement time points to the time interval. In order to improve the measurement accuracy, multiple sensors can be used for multi-point measurement, and then the average value can be taken or data fusion processing can be performed. At the same time, the sensor should be cleaned and maintained to prevent dust from contaminating the lens and affecting the measurement accuracy.

[0091] For the ash density, specific heat capacity at constant pressure and temperature change, ash samples are collected from the economizer regularly and then analyzed and measured in the laboratory. For the ash density, the pycnometer method or other density measuring instruments can be used for measurement. A certain mass of ash sample is placed in a container of known volume, and the ash density is obtained by measuring the ratio of mass to volume. The specific heat capacity at constant pressure of ash can be measured by thermal analysis techniques such as differential scanning calorimetry (DSC). During the measurement process, the ash sample is heated together with the standard material, and the specific heat capacity at constant pressure of ash is calculated by measuring the difference in heat change between the two during the heating process. The temperature change of ash can be measured by installing temperature sensors such as micro-thermocouples or thermistors in the ash layer, monitoring the temperature change inside the ash in real time during the ash accumulation process, and calculating the temperature difference at different time points to obtain the temperature change of ash. However, this experimental sampling and analysis method can only obtain the ash parameters at discrete time points, and needs to be combined with other online monitoring methods (such as ash thickness change rate measurement) to fully understand the dynamic changes of ash accumulation. At the same time, attention should be paid to the representativeness and accuracy of the sampling process to ensure that the collected ash samples can reflect the true characteristics of the ash accumulation in the economizer.

[0092] The application principles of the above-mentioned sensors and experiments are all existing technologies, so they will not be elaborated in detail.

[0093] In order to further optimize the above embodiment, in step 2, the temperature change rate model is:

[0094]

[0095] Where V is the effective volume of the economizer, ρ is the fluid density, C p is the specific heat capacity of the fluid at constant pressure, is the rate of change of temperature over time;

[0096]

[0097] Where ΔQ loss is the actual heat loss caused by ash accumulation, ΔT is the temperature difference between the inlet and outlet of the economizer, and H is the volume flow rate of the fluid;

[0098] Temperature change rate model Represents the rate of temperature change within the entire effective volume.

[0099] The convection heat transfer model is:

[0100] hA(T fluid -T wall )

[0101] Where h is the convective heat transfer coefficient, A is the heat transfer area in the economizer, T fluid is the economizer inlet fluid temperature, T wall is the economizer wall temperature;

[0102] Convective heat transfer model hA(T fluid -T wall ) represents the heat transfer process caused by the macroscopic movement of the fluid.

[0103] The radiation heat transfer model is:

[0104]

[0105] Where, ∈ is the wall emissivity, σ is the Stefan-Boltzmann constant, A is the heat transfer area in the economizer, T wall is the economizer wall temperature, T sur is the absolute temperature of the surrounding environment;

[0106] Radiation Heat Transfer Model Represents the heat exchange process caused by emitting and absorbing radiation energy.

[0107] The reference heat loss model is:

[0108]

[0109] in, is the dust accumulation mass flow rate, C p，ash is the constant pressure specific heat capacity of the ash, ΔT ash is the change of dust accumulation temperature;

[0110]

[0111] Where E is the ash thickness change rate, A is the heat exchange area in the economizer, ρ ash is the dust accumulation density;

[0112] Reference heat loss model An idealized process representing the change in heat transfer due to the presence of dust deposits.

[0113] It should be noted that the ambient absolute temperature is obtained based on the ambient temperature, and the specific calculation principle is prior art and will not be elaborated on;

[0114] The convection heat transfer model represents the heat transfer process caused by the macroscopic movement of the fluid; particles (molecules or atomic aggregates) in the fluid contact the wall during the flow process, transferring the heat they carry to the wall or obtaining heat from the wall;

[0115] The convective heat transfer coefficient is a key parameter that comprehensively reflects the influence of the flow state (laminar or turbulent) of the fluid, the physical properties of the fluid (such as thermal conductivity, density, specific heat capacity, etc.), the geometric shape and roughness of the wall, etc. on the convective heat transfer; for example, in the case of turbulent flow, the mixing of fluid particles is more intense, and heat can be transferred more effectively, so the convective heat transfer coefficient will be larger than that in laminar flow;

[0116] The radiation heat transfer model describes the heat exchange process generated by an object emitting and absorbing radiation energy in the form of electromagnetic waves (mainly infrared rays); all objects with a temperature above absolute zero will emit thermal radiation, and the amount of radiation energy is related to factors such as the object's temperature and surface properties (expressed by emissivity);

[0117] Emissivity indicates the ability of an object's surface to emit thermal radiation, and its value range is 0-1. The higher the emissivity, the stronger the ability of the object's surface to emit thermal radiation. For example, the emissivity of a black rough surface is close to 1, while the emissivity of a shiny metal surface is lower.

[0118] The Stefan-Boltzmann constant is a natural constant that determines the fundamental quantitative relationship between thermal radiation energy and temperature; it is a known value and therefore will not be elaborated on;

[0119] The heat loss model caused by dust accumulation represents the change in heat transfer caused by the presence of dust accumulation; dust accumulation adheres to the heat exchange surface, changing the original heat exchange process between the fluid and the wall, which is also the heat loss under ideal conditions.

[0120] In order to further optimize the above embodiment, in step three, the association model is:

[0121]

[0122] The model is combined based on the heat change rate ΔQ in the economizer; That is, generate a correlation model;

[0123] Substitute the key parameters into the correlation model and solve the actual heat loss ΔQ caused by dust accumulation at the current moment. loss .

[0124] It should be noted that the association model integrates the effects of convective heat transfer, radiation heat transfer and ash accumulation into one formula, and links these factors through the heat change rate. This allows us to no longer view each heat transfer mechanism or ash accumulation effect in isolation when analyzing the heat transfer of the economizer, but to consider the interaction between them. This allows soot blowing control to be dynamically adjusted according to the actual operating status, rather than based on a fixed time interval or a simple threshold. For example, when operating conditions such as load changes and fuel quality fluctuations change, the relevant parameters in the model (such as fluid temperature, flow rate, ash accumulation rate, etc.) will change accordingly. Through the calculation of the association model, changes in heat loss caused by ash accumulation can be discovered in time, thereby advancing or delaying the soot blowing operation, optimizing the soot blowing frequency and duration, and improving the operating efficiency and reliability of the economizer.

[0125] In order to further optimize the above embodiment, in step 5, the machine learning model is:

[0126]

[0127] Among them, f new is the adjusted soot blowing frequency, f0 is the current soot blowing frequency, ΔQ loss is the actual heat loss caused by dust accumulation at the current moment; ΔQ th is the preset reference heat loss threshold, and when ΔQ loss Exceed ΔQ th When , it means that the soot blowing frequency needs to be increased; k1 and k2 are both frequency adjustment coefficients, k1 is used to reflect the sensitivity of the situation when the severity of dust accumulation exceeds the threshold, and k2 is used to reflect the degree of reducing the soot blowing frequency when the dust accumulation is relatively light;

[0128] t newis the adjusted soot blowing time, t0 is the soot blowing time at the current moment, t max is the maximum soot blowing time; k3 is the time adjustment coefficient, which is used to reflect the quantitative degree of the relationship between the heat loss caused by soot accumulation and the soot blowing time;

[0129] And when t new <t min When t new =t min ;t min It is the minimum soot blowing time.

[0130] It should be noted that f new The principle of the model is that when the heat loss caused by soot accumulation exceeds the threshold, the soot blowing frequency is increased; when the heat loss is lower than the threshold, the soot blowing frequency is appropriately reduced to reduce unnecessary soot blowing operations; when it is equal to the threshold, the current soot blowing frequency is maintained.

[0131] t new The principle of the model is to adjust the soot blowing time according to the heat loss caused by soot accumulation. The greater the heat loss, the longer the soot blowing time will be, but it will not exceed the maximum soot blowing time; at the same time, the soot blowing time will not be less than the minimum soot blowing time to ensure the soot blowing effect and equipment safety.

[0132] Through these two models, the frequency and duration of soot blowing can be automatically adjusted according to the actual heat loss caused by soot accumulation, while ensuring the heat exchange efficiency of the low-temperature economizer, reducing equipment wear and achieving balanced operation. It should be noted that the parameters k1, k2, k3, ΔQ in the formula th ,t max ,t min It needs to be determined through experiments and debugging based on the specific model, operating conditions and other factors of the low-temperature economizer. The means are existing technologies and will not be elaborated on here.

[0133] In order to further optimize the above embodiment, the training process of the frequency adjustment coefficient and the time adjustment coefficient includes:

[0134] Data collection: Collect several sets of historical operating data of low-temperature economizers, including actual heat loss caused by ash accumulation under different operating conditions, soot blowing frequency, soot blowing duration, and corresponding equipment operating effects; equipment operating effects include changes in heat exchange efficiency and equipment wear;

[0135] Establish an objective function; the goal is to find a set of coefficients k1, k2, k3, so that after adjusting the soot blowing frequency and duration, the equipment can minimize equipment wear while ensuring heat exchange efficiency, and the frequency and duration of soot blowing operations are relatively reasonable;

[0136] Train to get the ideal coefficients; use the gradient descent method to calculate the partial derivatives of the objective function with respect to the coefficients k1, k2, and k3, and then update the parameters in the opposite direction of the gradient; through continuous iterations, until the objective function converges or reaches the preset number of iterations, the final trained coefficients k1, k2, and k3 are output.

[0137] It should be noted that, in one embodiment, first, a large amount of low-temperature economizer operation data needs to be collected. These data include actual heat loss caused by soot accumulation under different operating conditions (such as different loads, different fuel qualities, different environmental conditions, etc.), soot blowing frequency, soot blowing duration, and corresponding equipment operation effects (such as changes in heat exchange efficiency, equipment wear, etc.).

[0138] Define a comprehensive evaluation function, for example:

[0139] J=α*Heat-β*Equipment-γ*BlowingF-δ*BlowingD

[0140] Among them, α, β, γ, and δ are weight coefficients used to balance the importance of different factors. Heat is the heat exchange efficiency, Equipment is the degree of equipment wear (which can be indirectly measured by indicators such as vibration level and component life), BlowingF is the soot blowing frequency, and BlowingD is the soot blowing duration;

[0141] For the gradient descent method, the partial derivatives of the objective function J with respect to the coefficients k1, k2, and k3 are calculated, and then the parameters are updated in the opposite direction of the gradient; in each iteration, the update formula is as follows:

[0142]

[0143] Where n is the number of iterations, η is the preset learning rate; through continuous iterations, until the objective function converges or reaches the preset number of iterations, the final trained coefficients k1, k2, and k3 are output.

[0144] In order to further optimize the above embodiment, it also includes using a PID controller according to the adjusted soot blowing frequency f new and the adjusted sootblowing time t new Corresponding control of the action of the sootblowing equipment.

[0145] It should be noted that the algorithm of the PID controller includes proportional coefficient, integral coefficient and differential coefficient;

[0146] According to the dynamic characteristics of the low-temperature economizer and the response requirements for soot blowing control, the appropriate proportional coefficient is determined through experiments and simulations. The proportional coefficient determines the degree of response of the controller to the current error (the difference between the actual heat loss and the target heat loss). For example, if the proportional coefficient is large, the controller will produce a large output response to the error, allowing the soot blowing equipment to adjust its action quickly, but it may cause the system to overshoot; if the proportional coefficient is too small, the response speed is slow, and the ash accumulation may not be controlled in a timely and effective manner. In the initial debugging stage, the proportional coefficient can be gradually increased to observe the response of the system until a suitable value is found that can ensure the response speed and avoid excessive overshoot.

[0147] The integral coefficient is used to eliminate the steady-state error of the system. In sootblowing control, the steady-state error may be manifested as the actual heat loss caused by soot accumulation cannot be stabilized within the target range even if the sootblowing equipment works at a certain frequency and duration. Through the integral action, the controller accumulates past errors and gradually adjusts the output so that the system eventually reaches a stable state. The determination of the integral coefficient needs to consider the inertia and integral saturation of the system. If the integral coefficient is too large, the integral term may accumulate too quickly in the early stage of the system response, causing overshoot; if the integral coefficient is too small, it will take too long to eliminate the steady-state error. You can start with a smaller value and gradually increase the integral coefficient while monitoring the system response to ensure that the system remains stable while eliminating the steady-state error.

[0148] The differential coefficient is mainly used to predict the changing trend of the system and adjust the control output in advance. In the low-temperature economizer soot blowing control, the differential action can adjust the action of the soot blowing equipment according to the rate of change of heat loss caused by ash accumulation. For example, when it is detected that the ash accumulation rate suddenly accelerates (the rate of change of heat loss increases), the differential term will cause the controller to output a larger signal, prompting the soot blowing equipment to respond faster to prevent excessive accumulation of ash. The determination of the differential coefficient requires caution because the differential is sensitive to noise. Excessively large differential coefficients may amplify measurement noise and cause unstable controller output. Usually, the differential coefficient can be set to a smaller value first, and then fine-tuned according to the dynamic response of the system.

[0149] The control accuracy can be improved by adjusting the frequency and time through the PID controller.

[0150] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A low-temperature economizer soot blowing control method based on machine learning, characterized in that: include: Step 1: Obtain key parameters in the soot blowing process of the low-temperature economizer; the key parameters include fluid parameters, temperature parameters, heat exchange parameters and ash accumulation parameters; Step 2: Establishing independent models based on key parameters; the independent models include temperature change rate model, convection heat transfer model, radiation heat transfer model, and reference heat loss model; Step 3: Correlate the independent models based on the heat change rate in the economizer to obtain a correlation model; Step 4, solving the correlation model to determine the actual heat loss caused by dust accumulation at the current moment; Step five: Establish a machine learning model to adjust the soot blowing frequency and time based on the actual heat loss to ensure the heat exchange efficiency of the low-temperature economizer and achieve balanced operation.

2. The low-temperature economizer soot blowing control method based on machine learning according to claim 1 is characterized in that: In step 1, the key parameters include: The fluid parameters include the fluid density in the economizer, the fluid specific heat capacity at constant pressure, the effective volume of the economizer, and the volume flow rate of the fluid, which are obtained based on the pre-installed sensor equipment; Temperature parameters include economizer inlet fluid temperature, outlet fluid temperature, economizer wall temperature and ambient temperature, which are obtained based on pre-installed sensor equipment; The heat transfer parameters include the heat transfer area, convective heat transfer coefficient and wall emissivity in the economizer, which are determined based on experiments; The convective heat transfer coefficient represents the influence of the fluid flow state, physical properties, wall geometry and roughness on the convective heat transfer; The wall emissivity represents the ability of the surface of an object to emit thermal radiation, and its value range is {0, 1}; The ash accumulation parameters include the ash thickness change rate, ash density, constant pressure specific heat capacity of the ash and ash temperature change in the economizer, which are based on pre-installed sensor equipment and experimental measurements.

3. The low-temperature economizer soot blowing control method based on machine learning according to claim 2 is characterized in that: In step 2, the temperature change rate model is: Where V is the effective volume of the economizer, ρ is the fluid density, C p is the specific heat capacity of the fluid at constant pressure, is the rate of change of temperature over time; Where ΔQ loss is the actual heat loss caused by ash accumulation, ΔT is the temperature difference between the inlet and outlet of the economizer, and H is the volume flow rate of the fluid; The temperature change rate model Represents the rate of temperature change within the entire effective volume.

4. The low-temperature economizer soot blowing control method based on machine learning according to claim 3 is characterized in that: In step 2, the convection heat transfer model is: hA(T fluid -T wall ) Where h is the convective heat transfer coefficient, A is the heat transfer area in the economizer, T fluid is the economizer inlet fluid temperature, T wall is the economizer wall temperature; The convection heat transfer model hA(T fluid -T wall ) represents the heat transfer process caused by the macroscopic movement of the fluid.

5. The low-temperature economizer soot blowing control method based on machine learning according to claim 4 is characterized in that: In step 2, the radiation heat transfer model is: Where, ∈ is the wall emissivity, σ is the Stefan-Boltzmann constant, A is the heat transfer area in the economizer, T wall is the economizer wall temperature, T sur is the absolute temperature of the surrounding environment; The radiation heat transfer model Represents the heat exchange process caused by emitting and absorbing radiation energy.

6. The low-temperature economizer soot blowing control method based on machine learning according to claim 5 is characterized in that: In step 2, the reference heat loss model is: in, is the dust accumulation mass flow rate, C p , ash is the constant pressure specific heat capacity of the ash, ΔT ash is the change of dust accumulation temperature; Where E is the ash thickness change rate, A is the heat exchange area in the economizer, ρ ash is the dust accumulation density; The reference heat loss model An idealized process representing the change in heat transfer due to the presence of dust deposits.

7. The low-temperature economizer soot blowing control method based on machine learning according to claim 6 is characterized in that: In step three, the association model is: The model is combined based on the heat change rate Δq in the economizer; That is, generate a correlation model; Substitute the key parameters into the correlation model and solve the actual heat loss ΔQ caused by dust accumulation at the current moment. loss .

8. The low-temperature economizer soot blowing control method based on machine learning according to claim 7 is characterized in that: In step 5, the machine learning model is: Among them, f new is the adjusted soot blowing frequency, f0 is the current soot blowing frequency, ΔQ loss is the actual heat loss caused by dust accumulation at the current moment; ΔQ th is the preset reference heat loss threshold, and when ΔQ loss Exceed ΔQ th When , it means that the soot blowing frequency needs to be increased; k1 and k2 are both frequency adjustment coefficients, k1 is used to reflect the sensitivity of the situation when the severity of dust accumulation exceeds the threshold, and k2 is used to reflect the degree of reducing the soot blowing frequency when the dust accumulation is relatively light; t new is the adjusted soot blowing time, t0 is the soot blowing time at the current moment, t max is the maximum soot blowing time; k3 is the time adjustment coefficient, which is used to reflect the quantitative degree of the relationship between the heat loss caused by soot accumulation and the soot blowing time; And when t new <t min When t new =t min ;t min is the minimum soot blowing time.

9. The low-temperature economizer soot blowing control method based on machine learning according to claim 8, characterized in that: The training process of the frequency adjustment coefficient and the time adjustment coefficient includes: Data collection: Collect several sets of historical operating data of low-temperature economizers, including actual heat loss caused by ash accumulation under different operating conditions, soot blowing frequency, soot blowing duration, and corresponding equipment operating effects; the equipment operating effects include changes in heat exchange efficiency and equipment wear; Establish an objective function; the goal is to find a set of coefficients k1, k2, k3, so that after adjusting the soot blowing frequency and duration, the equipment can minimize equipment wear while ensuring heat exchange efficiency, and the frequency and duration of soot blowing operations are relatively reasonable; Train to get the ideal coefficients; use the gradient descent method to calculate the partial derivatives of the objective function with respect to the coefficients k1, k2, and k3, and then update the parameters in the opposite direction of the gradient; through continuous iterations, until the objective function converges or reaches the preset number of iterations, the final trained coefficients k1, k2, and k3 are output.

10. The low-temperature economizer soot blowing control method based on machine learning according to claim 9, characterized in that: It also includes the soot blowing frequency f adjusted by the PID controller new and the adjusted sootblowing time t new Corresponding control of the action of the sootblowing equipment.

Images