Power plant gas-fired boiler control method based on machine learning model prediction
Through machine learning models, the overheated steam flow changes of power plant gas boilers are predicted, and key control parameters are automatically adjusted, which solves the problem of low efficiency of traditional manual control and achieves efficient and stable boiler operation and energy conservation and emission reduction.
Patent Information
- Application Number
- CN202411062209.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-05
- Publication Date
- 2024-11-19
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The control of gas boilers in traditional power plants relies on manual experience, resulting in low combustion efficiency, untimely intervention, and inability to effectively adjust the combustion-assisted air flow, resulting in gas waste and a decline in process indicators, and lack of methods to predict future control strategies.
Using a prediction method based on machine learning model, an overheated steam flow prediction model is established, and the model is trained through historical data to predict the future steam flow change trend, and the smoke residual oxygen content, furnace negative pressure, boiler water level, and superheated steam are adjusted according to the prediction results. Pressure and temperature achieve automatic control.
Real-time and precise control of the boiler is achieved, steam efficiency and process stability are improved, electricity and water consumption for auxiliary machines are reduced, and the savings of the combustion system are improved.
Smart Images

Figure CN118980087A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power plant gas-fired boiler technology, and specifically to a power plant gas-fired boiler control method based on machine learning model prediction. Background Technology
[0002] Gas-fired boilers in power plants utilize the heat released from burning surplus blast furnace gas, converter gas, coke oven gas, and other fuels from steel plants to heat water or other working media, thereby obtaining steam, hot water, or other working media with specified parameters (temperature, pressure) and quality. In industrial enterprises, gas-fired boilers in power plants primarily provide high-temperature, high-pressure steam to steam turbine generator sets for power generation.
[0003] Gas-fired boilers in power plants operate continuously. The main components of a gas-fired boiler include: a steam drum, furnace and combustion equipment, superheaters (low-temperature superheaters, screen-type superheaters, high-temperature superheaters), reheaters (low-temperature reheaters, high-temperature reheaters), desuperheaters, economizers, air preheaters, and gas preheaters. Auxiliary equipment includes: feedwater pumps, forced draft fans, induced draft fans, desulfurization equipment, denitrification equipment, dust removal equipment, and chimneys. The main operating systems of a gas-fired boiler in a power plant are divided into the combustion system and the steam-water system.
[0004] In the combustion system, blast furnace gas, converter gas, and coke oven gas, delivered through gas pipelines and gas preheaters, along with hot air supplied by blowers, air preheaters, and air supply pipelines, enter the boiler furnace. Burners on the front and rear walls of the furnace mix the air and gas for complete combustion. The rapid combustion of air and gas within the furnace releases a large amount of heat. Numerous water-cooled pipes are arranged around the inner walls of the furnace, while high-temperature superheaters, screen-type superheaters, and low-temperature superheaters are located at the top of the furnace. The water-cooled walls and superheaters are the radiant heating surfaces of the furnace; the working fluid inside absorbs radiant heat from the furnace while simultaneously lowering the flame temperature, protecting the furnace walls from damage. The high-temperature flue gas generated by combustion in the furnace leaves the furnace through the upper outlet and enters the horizontal flue. It exchanges heat with the low-temperature reheater and high-temperature reheater arranged in the horizontal flue. Then it enters the tail flue and exchanges heat with the economizer, air preheater and gas preheater in sequence, so that the flue gas gradually cools down by continuously releasing heat. The low-temperature flue gas then passes through the dust collector, desulfurization equipment and denitrification equipment to remove most of the pollutants in the flue gas. Finally, only a small amount of pollutants are sent into the chimney by the induced draft fan and discharged into the atmosphere.
[0005] In the steam-water system, water pumped to the boiler by the feedwater pump first enters the economizer, where it absorbs heat from the boiler tail flue gas before entering the steam drum. It is then introduced into the cooling water lower header via the downcomer and distributed to the various water-cooled wall tubes. In the water-cooled walls, the water absorbs radiant heat from the high-temperature flames and flue gas in the furnace, causing some of the water to evaporate into saturated steam, thus forming a steam-water mixture within the water-cooled walls. This mixture flows upwards and enters the steam drum, where it is separated by a steam-water separator. The separated water continues to circulate. The separated saturated steam is sent to the superheater for heating via the saturated steam outlet pipe at the top of the steam drum. Finally, the superheated steam that meets the requirements is introduced into the turbine through the superheated steam pipeline to perform work. Besides steam-water separation, the steam drum also stores a certain amount of water, providing a certain amount of heat and working fluid reserves. This helps to slow down the rate of steam pressure change during operating conditions and acts as a buffer when the feedwater and load are not coordinated for a short period. In addition, the quality of the superheated steam decreases after it has done work in the turbine. This portion of the steam enters the reheater for reheating to improve its quality, allowing it to continue doing work in the turbine. Some superheated steam, after doing work in the turbine, becomes lower-quality exhaust steam. This exhaust steam is directly converted into water through the condenser and recycled. Finally, to better control the temperature of the superheated and reheated steam, desuperheating water supplied by a feedwater pump is introduced between the superheater and reheater to cool the steam, ensuring it meets the turbine's work requirements.
[0006] The control of gas-fired boilers in power plants mainly includes the following key circuits: steam drum water level control circuit, furnace negative pressure control circuit, flue gas residual oxygen content control circuit, superheated steam pressure control circuit, and steam temperature control circuit.
[0007] The steam drum water level control loop mainly controls the steam drum water level within the required parameter range by adjusting the frequency of the feedwater pump, the main feedwater regulating valve (feed volume), the emergency drain valve of the steam drum, and the continuous blowdown valve of the steam drum.
[0008] The furnace negative pressure control circuit mainly controls the gas flow rate of the burner, the air flow rate of the blower, and the induced draft fan, thereby controlling the gas flow rate and flue gas flow rate in the furnace and keeping the pressure in the furnace within the required parameter range.
[0009] The flue gas residual oxygen content control loop mainly controls the oxygen concentration in the flue gas within the required parameter range by adjusting the gas flow rate of the burner and the air flow rate of the blower.
[0010] The superheated steam pressure control loop mainly controls the superheated steam pressure and steam drum pressure within the reasonable range required by the turbine generator set by adjusting operating variables such as the gas flow rate of the burner, the air flow rate of the forced draft fan, the induced draft fan, and the water flow rate of the feedwater pump and feedwater regulating valve.
[0011] The main function of the steam temperature control loop is to control the superheated steam temperature and reheated steam temperature within the reasonable range required by the turbine generator set by adjusting the flow rates of superheated steam desuperheating water, reheated steam desuperheating water, and flue gas electric regulating damper valve (flue gas flow).
[0012] Traditional power plant gas-fired boiler control relies heavily on manual methods. Personnel monitor key operating parameters such as drum water level, furnace negative pressure, flue gas residual oxygen content, superheated steam flow rate, superheated steam pressure, and superheated steam temperature. Adjustments are made to burner gas valves, forced draft fan frequency, induced draft fan frequency, feedwater valves, feedwater pump frequency, and desuperheating water valves based on experience. Because manual control depends on experience, some personnel lack the expertise to optimize these key process parameters. Furthermore, manual operation often suffers from delayed intervention; fluctuations in operating conditions prevent immediate intervention, leading to gas waste and a decline in process parameters. Additionally, the control of combustion air flow currently employs a fixed air-to-gas ratio, multiplying the total gas flow rate by a fixed value. However, due to fluctuations in gas composition, a fixed air-to-gas ratio reduces combustion efficiency. Current methods typically only involve manual adjustment when the air-to-gas ratio is severely imbalanced, resulting in low efficiency and delayed intervention. In the field of existing industrial control, although there are schemes that use energy consumption data to predict future times to set corresponding control strategies, there is a gap in this aspect in the field of power plant gas boiler control, and there is no way to determine the ideal gas boiler control strategy for the current time. Summary of the Invention
[0013] To address the aforementioned technical problems, this invention provides a control method for gas-fired boilers in power plants based on machine learning model prediction.
[0014] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:
[0015] A method for controlling a gas-fired boiler in a power plant based on machine learning model prediction includes the following steps:
[0016] Step 1: Establish a predictive model for superheated steam flow rate based on a machine learning model:
[0017] A production dataset D1 is generated from historical production data of the power plant's gas-fired boilers. This production data includes the following variables: blast furnace gas pipeline flow rate, blast furnace gas pipeline pressure, furnace negative pressure, blower flow rate, residual oxygen content in flue gas, calorific value of blast furnace gas, air-to-coal ratio, turbine generator power output, and superheated steam flow rate. Superheated steam flow rate is the target variable, while the remaining variables are prediction variables used to train the prediction model. First, the production dataset D1 is clustered based on these variables, dividing it into N training classes. One prediction model is trained for each class, resulting in N prediction models. In actual production, the current production data is first classified to obtain the classification results. Based on these results, the current production data is input into the corresponding trained prediction model, which then predicts the superheated steam flow rate at p future moments.
[0018] Step 2: Determine the state of superheated steam flow rate at p future time points: The superheated steam flow rate at the current time t is S. t The standard deviation of the superheated steam flow rate over the past p time steps is σ. t-p The prediction model forecasts the superheated steam flow rate S for p future moments. t+1 ,S t+2 ,…,S t+p The change in superheated steam flow rate ΔS after p time points t+p =S t+p -S t ;
[0019] If |ΔS t+p |≤|σ t-p If |, then the superheated steam flow rate at the next p moments is considered to be in a steady state;
[0020] If |ΔS t+p |>|σ t-p |, then the superheated steam flow rate at the next p moments is considered to be in an unsteady state; and when ΔS t+p >0, σ t-p When ΔS > 0, the superheated steam flow rate will increase in the next p time steps; and when ΔS t+p <0, σ t-p When the value is less than 0, the superheated steam flow rate will decrease in the next p time intervals.
[0021] Step 3: Based on the state of superheated steam flow at p future moments, control the residual oxygen content in the flue gas, furnace negative pressure, steam drum water level, superheated steam pressure, superheated steam temperature, and reheated steam temperature at the current moment.
[0022] Furthermore, in step three, the control of residual oxygen content in flue gas, furnace negative pressure, steam drum water level, superheated steam pressure, superheated steam temperature, and reheated steam temperature specifically includes:
[0023] By adjusting the gas flow rate and air flow rate of the burner, the residual oxygen content in the flue gas can be controlled within a set range;
[0024] The negative pressure in the furnace is controlled within the set range by adjusting the induced draft fan flow rate.
[0025] By adjusting the main feedwater flow rate, the steam drum water level is controlled within the set range;
[0026] The superheated steam pressure is controlled within a set range by adjusting the gas flow rate of the burner, the air flow rate of the blower, and the frequency of the feed water pump.
[0027] The temperature of the superheated steam is controlled within the set range by adjusting the flow rate of the desuperheating water in the secondary desuperheater of the superheated steam.
[0028] The reheat steam temperature is controlled within the set range by adjusting the reheat steam desuperheating water flow rate and the flue gas flow rate.
[0029] Furthermore, the step of controlling the residual oxygen content in the flue gas within a set range by adjusting the gas flow rate and air flow rate of the burner specifically includes:
[0030] The lower limit for residual oxygen content in flue gas is set as o. low The upper limit for residual oxygen content in flue gas is o. high ;
[0031] If the superheated steam flow rate is in a steady state at p future moments, then the following judgment is made:
[0032] If the residual oxygen content in the flue gas is below 0 low If the residual oxygen content in the flue gas continues to decrease within the set time, then the airflow rate of the blower and the airflow rate of the burner are increased by increasing the frequency of the blower and increasing the given opening degree of the burner air valve, until the residual oxygen content in the flue gas is not lower than 0. low ;
[0033] If the residual oxygen content in the flue gas is higher than o high If the residual oxygen content in the flue gas continues to rise within the set time, the blower's airflow and the burner's airflow will be reduced by decreasing the frequency of the blower and decreasing the given opening of the burner's air valve, until the residual oxygen content in the flue gas does not exceed 0. high ;
[0034] If the superheated steam flow rate is about to increase at future p time points, then first increase the gas flow rate and the air flow rate of the burner: This requires first increasing the given opening degree of the burner gas inlet valve, then increasing the given frequency of the blower's variable frequency drive, and finally increasing the given opening degree of the burner air valve; then the following judgments are made:
[0035] If the residual oxygen content in the flue gas is below 0 low If the residual oxygen content in the flue gas continues to decrease within the set time, then increase the air-to-coal ratio and increase the air flow rate of the blower. In this case, first increase the frequency of the blower's inverter, then increase the given opening degree of the burner air valve until the residual oxygen content in the flue gas is not lower than 0. low ;
[0036] If the residual oxygen content in the flue gas is higher than o high If the residual oxygen content in the flue gas continues to rise within the set time, reduce the air-coal ratio and reduce the air flow rate of the blower: at this time, the frequency of the blower should be reduced first, and then the air valve of the burner should be reduced to the given opening degree.
[0037] If the superheated steam flow rate is about to decrease at p future moments, then first reduce the gas flow rate and then the air flow rate of the burner: This requires first reducing the given opening of the burner gas inlet valve, then reducing the given frequency of the blower's frequency converter, and finally reducing the given opening of the burner air valve; then the following judgments are made:
[0038] If the residual oxygen content in the flue gas is below 0 low If the residual oxygen content in the flue gas continues to decrease within the set time, then increase the air-to-coal ratio and increase the air flow rate of the blower. In this case, first increase the frequency of the blower's inverter, then increase the given opening degree of the burner air valve until the residual oxygen content in the flue gas is not lower than 0. low ;
[0039] If the residual oxygen content in the flue gas is higher than o high If the residual oxygen content in the flue gas continues to rise within the set time, reduce the air-to-coal ratio and decrease the air flow rate of the blower. In this case, first reduce the frequency of the blower's inverter, then reduce the given opening of the burner air valve until the residual oxygen content in the flue gas is no higher than 0. high .
[0040] Furthermore, the step of controlling the furnace negative pressure within a set range by adjusting the induced draft fan flow rate specifically includes:
[0041] Let p be the set lower limit of the furnace negative pressure. low The upper limit of the negative pressure in the furnace is p high ;
[0042] If the superheated steam flow rate is in a steady state at p future moments, then the following judgment is made:
[0043] If the furnace negative pressure is higher than p high Then, the induced draft fan's airflow is increased by increasing its variable frequency drive until the furnace negative pressure does not exceed p. high If the furnace negative pressure is lower than p lowThen, the induced draft fan's airflow is reduced by decreasing the frequency of the induced draft fan's inverter until the furnace negative pressure is not lower than p. low ;
[0044] If the superheated steam flow rate is about to increase at future p time points, then first increase the gas flow rate and the air flow rate of the burner: This requires first increasing the given opening degree of the burner gas inlet valve, then increasing the given frequency of the blower's variable frequency drive, and finally increasing the given opening degree of the burner air valve; then the following judgments are made:
[0045] If the furnace negative pressure is higher than p high Then, the induced draft fan's airflow is increased by increasing its variable frequency drive until the furnace negative pressure does not exceed p. high ;
[0046] If the furnace negative pressure is lower than p low Then, the induced draft fan's airflow is reduced by decreasing the frequency of the induced draft fan's inverter until the furnace negative pressure is not lower than p. low ;
[0047] If the superheated steam flow rate is about to decrease at p future moments, then first reduce the gas flow rate and then the air flow rate of the burner: This requires first reducing the given opening of the burner gas inlet valve, then reducing the given frequency of the blower's frequency converter, and finally reducing the given opening of the burner air valve; then the following judgments are made:
[0048] If the furnace negative pressure is higher than p high Then, the induced draft fan's airflow is increased by increasing its variable frequency drive until the furnace negative pressure does not exceed p. high ;
[0049] If the furnace negative pressure is lower than p low Then, the induced draft fan's airflow is reduced by decreasing the frequency of the induced draft fan's inverter until the furnace negative pressure is not lower than p. low .
[0050] Furthermore, the step of controlling the steam drum water level within a set range by adjusting the main feedwater flow rate specifically includes:
[0051] If the superheated steam flow rate is stable at p future moments, the main feedwater regulating valve is adjusted according to the real-time drum water level: the trend of the drum water level is judged; if the drum water level shows a downward trend, the opening of the main feedwater regulating valve is increased; if the drum water level shows an upward trend, the opening of the main feedwater regulating valve is decreased. The pressure difference between the feedwater pump downstream header pressure and the drum pressure is monitored, and the drum water level and the pressure difference between the feedwater pump downstream header pressure and the drum pressure are controlled within the set range by adjusting the frequency of the feedwater pump.
[0052] If the superheated steam flow rate is about to increase at p future moments, the water level in the steam drum will drop and the steam drum pressure will increase. At this time, the main feedwater regulating valve of the boiler will be opened to increase the main feedwater flow rate in the steam drum and raise the water level in the steam drum. The frequency of the feedwater pump will be increased by a given value to increase the feedwater flow rate and the pressure of the feedwater pump downstream header until the superheated steam flow rate reaches a stable state and the pressure difference between the steam drum water level, the feedwater pump downstream header pressure and the steam drum pressure is within the set range.
[0053] If the superheated steam flow rate is about to decrease at p future moments, the water level in the steam drum will rise and the steam drum pressure will decrease. At this time, the boiler main feedwater regulating valve will be reduced by a given opening to reduce the main feedwater flow rate in the steam drum and lower the steam drum water level. The frequency converter of the feedwater pump will be reduced by a given value to reduce the feedwater flow rate and the pressure of the feedwater pump downstream header until the superheated steam flow rate reaches a stable state and the pressure difference between the steam drum water level, the feedwater pump downstream header pressure, and the steam drum pressure is within the set range.
[0054] Furthermore, the control of superheated steam pressure within a set range by adjusting the gas flow rate of the burner, the air flow rate of the blower, and the frequency conversion frequency of the feedwater pump specifically includes:
[0055] If the superheated steam flow rate is in a steady state at p future moments, then there is no need to adjust the superheated steam pressure.
[0056] If the superheated steam flow rate is about to increase at p future moments, firstly, determine whether it is necessary to increase the main feedwater flow rate into the steam drum based on the current steam drum water level: Let the set upper limit of the steam drum water level be w. high The lower limit of the steam drum water level is w low If the current steam drum water level is higher than w high In this case, it is not necessary to adjust the main feedwater flow rate; if the current steam drum water level is higher than 0mm and lower than w high This will increase the main water supply flow rate by |ΔS t+p If the steam drum water level is below 0mm, first increase the main feedwater flow rate until the steam drum water level reaches 0mm, then increase the main feedwater flow rate by |ΔS t+p i. Secondly, monitor the pressure difference between the feedwater pump downstream header pressure and the steam drum pressure, adjust the feedwater pump frequency converter frequency, and control the steam drum water level and the pressure difference between the feedwater pump downstream header pressure and the steam drum pressure within the set range; finally, while maintaining the furnace negative pressure and the blast furnace gas pipeline pressure within the process requirements, increase the opening of the burner gas regulating valve until the superheated steam flow reaches the superheated steam flow predicted by the prediction model at a specific time. At the same time, increase the blower air flow to control the residual oxygen content in the flue gas within the set range.
[0057] If the superheated steam flow rate is about to decrease at p future moments, first determine whether it is necessary to reduce the main feedwater flow rate into the steam drum based on the current steam drum water level: Let w be the set upper limit of the steam drum water level. high The lower limit of the steam drum water level is w low If the current steam drum water level is higher than w high At this point, first adjust the steam drum water level to w. high Then reduce the main water supply flow rate by |ΔS t+p |;If the current steam drum water level is higher than 0mm and lower than w high This will reduce the main water supply flow rate by |ΔS t+p If the current steam drum water level is below 0mm, increase the main feedwater flow rate until the steam drum water level reaches 0mm, then increase the main feedwater flow rate by |ΔS t+p Secondly, monitor the pressure difference between the feedwater pump's downstream main pipe pressure and the steam drum pressure, adjust the feedwater pump's frequency converter frequency, and control the steam drum water level and the pressure difference between the feedwater pump's downstream main pipe pressure and the steam drum pressure within the set range. Finally, while maintaining the furnace negative pressure and the blast furnace gas pipeline pressure within the process requirements, reduce the opening of the burner's gas regulating valve until the superheated steam flow rate reaches the superheated steam flow rate predicted by the prediction model at a specific time. At the same time, reduce the blower's air flow rate to control the residual oxygen content in the flue gas within the set range.
[0058] Furthermore, the step of controlling the superheated steam temperature within a set range by adjusting the flow rate of the desuperheating water in the secondary superheated steam desuperheater specifically includes:
[0059] Using the superheated steam temperature after the outlet header of the screen-type superheater as feedforward information and the superheated steam temperature after the outlet header of the high-temperature superheater as the control target, the superheated steam temperature is controlled by adjusting the opening of the regulating valve of the superheated steam secondary desuperheater.
[0060] Based on the principle of energy balance, the working fluid energy is balanced before and after the superheated steam desuperheater at the outlet header of the screen-type superheater. The working fluid includes steam and water. Therefore, the energy of the steam pipe before the desuperheating water is L. s ×H s (P s ,T s )+L w ×H w (P w ,T w ); where L s H s P s T s These are the superheated steam flow rate, superheated steam enthalpy, superheated steam pressure, and superheated steam temperature before the secondary superheated steam desuperheater; L w H wP w T w These are the desuperheating water flow rate, desuperheating water enthalpy, desuperheating water pressure, and desuperheating water temperature of the secondary desuperheating unit for superheated steam.
[0061] The energy of the steam pipeline after the superheated steam secondary desuperheater is L. out ×H out (P out ,T out ); where L out H out P out T out These are the superheated steam flow rate, superheated steam enthalpy, superheated steam pressure, and superheated steam temperature after the second-stage superheated steam desuperheater;
[0062] Energy balance formula:
[0063] L s ×H s (P s ,T s )+L w ×H w (P w ,T w )=: out ×H out (P out ,T out );
[0064] Among them, the desuperheating water flow rate L of the superheated steam secondary desuperheater w The unknown quantity is the temperature T of the superheated steam after the secondary desuperheater. out The superheated steam temperature is an indirect target temperature. There is a correlation between the superheated steam temperature and the superheated steam flow rate. The control target temperature of the superheated steam temperature is determined by combining the predicted superheated steam flow rate with the boiler load characteristic curve.
[0065] The desuperheating water flow rate L of the secondary desuperheater of superheated steam is derived using the energy balance formula. w :
[0066]
[0067] Will: w The value is used as the target value, and the flow rate of the desuperheating water is adjusted by the superheater desuperheating water regulating valve to control the superheated steam temperature within the set range.
[0068] Furthermore, the step of controlling the reheat steam temperature within a set range by adjusting the reheat steam desuperheating water flow rate and the flue gas flow rate specifically includes:
[0069] There is a relatively definite correlation between reheat steam temperature and superheat steam flow rate. The target temperature for reheat steam temperature is determined by combining the predicted superheat steam flow rate with the boiler load characteristic curve.
[0070] By monitoring the reheat steam temperature after the high-temperature reheater and the rate of change of the reheat steam temperature after the high-temperature reheater, the opening of the flue gas electric regulating damper valve is adjusted using a PID algorithm to control the reheat steam temperature within the set range.
[0071] Furthermore, the prediction model employs the LightGBM regression model.
[0072] Compared with the prior art, the beneficial technical effects of the present invention are:
[0073] This invention combines current and past data on the flow and pressure of blast furnace gas pipelines in power plants, key operating indicators of gas-fired boilers, and changes in the power generation load of turbine generator units. Through a machine learning model, it predicts the trend of boiler steam output over a future time period. Based on the changing trend of superheated steam flow, it allows for advance prediction of control operations in key boiler loops, enabling timely and accurate adjustments. Simultaneously, machine learning utilizes extensive historical data to process and analyze the complex operating conditions of the boiler, providing significant advantages over other automatic control methods. First, this invention better captures the temporal characteristics of data, adapting to changes in equipment operating conditions. Second, this method is largely independent of manual intervention, enabling real-time and precise control of key systems such as the boiler's combustion system (gas volume, forced draft flow, induced draft flow), steam generation system (superheated steam pressure, superheated steam temperature), and feedwater system (feedwater flow). This improves boiler steam efficiency and conserves key working fluids (auxiliary power consumption, water consumption, environmental chemicals, etc.), while providing more stable control over the process indicators of each key loop in the power plant's gas-fired boiler. Attached Figure Description
[0074] Figure 1 This is a schematic diagram of the overall process of the present invention.
[0075] Figure 2 A schematic diagram for constructing hysteresis features. Detailed Implementation
[0076] A preferred embodiment of the present invention will now be described in detail with reference to the accompanying drawings.
[0077] This invention provides a control method for gas-fired boilers in power plants based on machine learning model prediction. Figure 1 The overall process of this invention has been simply illustrated. The following sections will provide a detailed description of the process in several parts.
[0078] 1. Establish a predictive model for superheated steam flow rate based on a machine learning model:
[0079] The gas-fired boilers in steel plant self-owned power plants, also known as "residual gas boilers," refer to boilers that burn blast furnace gas to produce high-temperature, high-pressure steam. This gas comes from the blast furnace gas produced after other main gas-consuming equipment in the steel plant has been supplied with the gas. The gas-fired boilers convert this surplus blast furnace gas into electricity for the steel plant, saving energy. Therefore, the amount of blast furnace gas that the gas-fired boiler can burn is not determined by the power generation load at the power generation end, but by the blast furnace end and its main steam-consuming equipment; that is, the power plant burns only the amount of gas supplied by the blast furnace. This production mechanism of steel plant self-owned power plants leads to the following situation: to achieve energy saving and efficiency improvement, it is only possible to improve the boiler efficiency and the power generation efficiency of the turbine generator set. Since the power generation efficiency of the turbine generator set is generally quite high, it is not within the scope of this invention. Therefore, this paper only considers improving the boiler efficiency, that is, using the limited blast furnace gas supplied by the blast furnace to the gas-fired boilers of the self-owned power plant to produce as much steam as possible that meets the power generation requirements of the turbine generator set, thus generating as much electricity as possible.
[0080] To improve the overall operating efficiency of a boiler, it is necessary to coordinate and precisely control the boiler's combustion system, steam generation system, and feedwater system. To better control the aforementioned systems of the boiler, the future trend of superheated steam flow is crucial. Firstly, the trend directly reflects changes in the consumption of blast furnace gas, the primary fuel in the combustion system. Maintaining complete combustion of boiler fuel and keeping boiler negative pressure and residual oxygen content in flue gas within the required process range is essential, and these two key indicators are directly related to the forced draft fan flow and induced draft fan flow. Secondly, changes in superheated steam temperature and pressure are directly affected by the superheated steam flow. An increase in superheated steam flow naturally leads to higher temperatures and pressures, and vice versa. If changes in superheated steam flow can be predicted in advance, controlling these temperatures and pressures becomes much simpler. Finally, changes in superheated steam flow are also closely related to the boiler feedwater system. A sudden increase in superheated steam flow directly affects the water level in the steam drum, causing it to drop. In this case, water should be added in advance to maintain the steam drum water level within a safe range, and vice versa.
[0081] In summary, the changing trend of superheated steam flow in a boiler is crucial for boiler operation control and efficiency improvement. Therefore, this invention establishes a predictive model for the superheated steam flow in a power plant over a future period based on historical data of relevant variables in the boiler combustion system and the boiler steam-water system. This model can accurately estimate the superheated steam flow and its changing trend over a future period. Specifically, it includes:
[0082] A production dataset D1 is generated from historical production data of the power plant's gas-fired boilers. This production data includes the following variables: blast furnace gas pipeline flow rate, blast furnace gas pipeline pressure, furnace negative pressure, blower air flow rate, residual oxygen content in flue gas, calorific value of blast furnace gas, air-to-coal ratio, turbine generator power generation, and superheated steam flow rate. Superheated steam flow rate is the target variable, while the remaining variables are prediction variables used to train the prediction model. First, the production dataset D1 is clustered based on these variables, dividing it into N training classes. One prediction model is trained for each class, resulting in a total of N prediction models. In actual production, the current production data is first classified. Based on the classification results, the production data is input into the corresponding trained prediction model, which then predicts the superheated steam flow rate at p future moments.
[0083] In a preferred embodiment, before using the production data from the production dataset to train the prediction model, lagged features are constructed using a sliding window method. A sliding window is used to select training samples from the acquired production dataset D1, with n time points as the window length. The window slides by one time point at a time, m times, and the data obtained from each sliding window is used as a new feature variable. For example, the following time windows [tn,t-(n-1),…,t-1], [t-(n-1),t-(n-2),…,t],…,[t,t+1,t+2,…,t+(n-1)], [t+1,t+2,…,t+n] are slid m times over the nine variable observations in the production dataset D1, effectively constructing m lagged feature variables for each variable, ultimately resulting in 9m feature variables for training the model. Figure 2 This demonstrates the characteristic variables of a variable constructed by sliding a window of length n through m observations, where... Figure 2 The table on the left shows the original time series variables. Figure 2 The right side of the graph represents m training data feature variables constructed using a sliding window. Finally, training sample selection is performed: from the training samples constructed using the above features, multiple continuous sequence data points where the residual oxygen content at each time point falls within a reasonable range required by the process are selected as the final training dataset. This ensures that the selected training model uses data with optimal combustion conditions.
[0084] The construction of lagging features of the production dataset and the classification of samples on the production dataset can be used together or independently as a preferred embodiment in this invention.
[0085] In a preferred embodiment, the prediction model employs the LightGBM regression model. LightGBM is an open-source ensemble learning machine learning model based on gradient boosting trees, released by Microsoft. Ensemble learning refers to building multiple learners (weak learners) using the training dataset, having each learner predict the test dataset separately, and then combining the predictions from all learners using a certain strategy to obtain the final prediction. Gradient boosting trees refer to the use of gradient boosting during model training. Each iteration of the model trains a new decision tree, and in subsequent training iterations, the performance of the previous decision tree is considered; this decision tree is a weak learner. The LightGBM regression model has actually been a particularly outstanding model in the field of machine learning in recent years. It offers significant improvements over previous models such as GBDT and XGBoost in terms of decision tree training strategies, internal feature engineering, and computer engineering optimization. Specifically, the LightGBM regression model is a comprehensive optimization of the XGBoost model. It employs a leaf-wise decision tree growth strategy (with maximum depth constraints) and uses a histogram algorithm (8 bits memory usage, lower time complexity) to accelerate tree generation by finding the optimal split point. It reduces the involvement of small gradient samples through the GOSS (Gradient One-Sided Sampling) algorithm and achieves feature dimensionality reduction by bundling mutually exclusive features using EFB. Furthermore, it automatically handles missing values and supports categorical features. The resulting improvements include faster model training speed, lower memory usage, and superior prediction performance. Therefore, the LightGBM regression model can effectively handle most prediction tasks, including the typical time series prediction task presented in this invention. Of course, during the training process of the LightGBM regression model, a lot of parameter tuning is required. For example, the "regression" parameter can be selected for time series prediction tasks, and a regressor with a regularization term can be selected. The MPAE (mean absolute percentage error) is more reasonable for time series data and can well evaluate the model performance on time series. Other parameters include the specific boosting algorithm, the limiting parameters for decision tree generation, and some engineering parameters for computer resource usage.
[0086] In a preferred embodiment, the production data under different load conditions in the production dataset D1 are clustered using the K-means clustering algorithm based on the superheated steam flow rate, dividing the production dataset D1 into N training samples. A prediction model is trained for each training sample, resulting in a total of N prediction models. In actual production, the current production data is first input into the K-means clustering model obtained after clustering based on the K-means clustering algorithm to obtain the category of the current training sample. Based on the classification result, the current production data is input into the corresponding completed prediction model, and the prediction model predicts the superheated steam flow rate at p future times.
[0087] Step 2, determine the current superheated steam flow rate: the current superheated steam flow rate is S. t The standard deviation of the superheated steam flow rate over the past p time steps is σ. t-p The prediction model forecasts the superheated steam flow rate S for p future moments. t+1 ,S t+2 ,…,S t+p The change in superheated steam flow rate ΔS after p time points t+p =S t+p -S t ;
[0088] If |ΔS t+p |≤|σ t-p If |, then the superheated steam flow rate at the next p moments is considered to be in a steady state;
[0089] If |ΔS t+p |>|σ t-p |, then the superheated steam flow rate at the next p moments is considered to be in an unsteady state; and when ΔS t+p >0, σ t-p When ΔS > 0, the superheated steam flow rate will increase in the next p time steps; and when ΔS t+p <0, σ t-p When the value is less than 0, the superheated steam flow rate will decrease in the next p time intervals.
[0090] Step 3: If the superheated steam flow rate at future p time points is in a steady state or an unstable state, then at the current time, based on the state of the superheated steam flow rate, control the residual oxygen content in the flue gas, the negative pressure in the furnace, the water level in the steam drum, the superheated steam pressure, the superheated steam temperature, and the reheated steam temperature.
[0091] Furthermore, in step three, the control of residual oxygen content in flue gas, furnace negative pressure, steam drum water level, superheated steam pressure, superheated steam temperature, and reheated steam temperature specifically includes:
[0092] By adjusting the gas flow rate and air flow rate of the burner, the residual oxygen content in the flue gas can be controlled within a set range;
[0093] The negative pressure in the furnace is controlled within the set range by adjusting the induced draft fan flow rate.
[0094] By adjusting the main feedwater flow rate, the steam drum water level is controlled within the set range;
[0095] The superheated steam pressure is controlled within a set range by adjusting the gas flow rate of the burner, the air flow rate of the blower, and the frequency of the feed water pump.
[0096] The temperature of the superheated steam is controlled within the set range by adjusting the flow rate of the desuperheating water in the secondary desuperheater of the superheated steam.
[0097] The reheat steam temperature is controlled within the set range by adjusting the reheat steam desuperheating water flow rate and the flue gas flow rate.
[0098] 2. Control loop for residual oxygen content in flue gas.
[0099] 2.1 The principle for controlling the residual oxygen content in flue gas is as follows:
[0100] The most important aspect of improving the efficiency of gas-fired boilers in power plants is enhancing the combustion efficiency of the boiler's fuel (blast furnace gas). To improve this efficiency, it's crucial to continuously adjust the air-to-gas ratio entering the furnace through the burners to achieve the optimal balance, ensuring complete combustion and maximizing heat release. The optimal air-to-gas ratio is reflected in the residual oxygen content or carbon monoxide content (not considered here) of the flue gas. Generally, a good residual oxygen content is controlled within the range of 1.5 to 2.5. Therefore, to improve boiler combustion efficiency, the residual oxygen content in the flue gas must be constantly adjusted and controlled within the process requirements to ensure complete fuel combustion within the furnace.
[0101] Since the fuel quantity (blast furnace gas flow) of a power plant's gas-fired boiler is determined by the blast furnace side and cannot be manually controlled, the flow rates of gas, forced draft, and induced draft entering the furnace can only be controlled while ensuring stable gas pipeline pressure. However, by obtaining the trend of superheated steam flow, the boiler's future combustion load can be relatively accurately predicted based on the past, present, and future changes in superheated steam flow, thereby adjusting the intake air, forced draft, and induced draft within the boiler furnace. The regulation and control of residual oxygen content in boiler flue gas is affected by the forced draft fan flow and the burner's gas flow. By predicting the superheated steam flow trend, the boiler's combustion load can be easily anticipated, allowing for advance adjustment of the burner's air flow, gas flow, and forced draft fan flow.
[0102] 2.2 The specific control logic for residual oxygen content in flue gas is as follows: By adjusting the gas flow rate and air flow rate of the burner, the residual oxygen content in the flue gas is controlled within a set range, specifically including:
[0103] The lower limit for residual oxygen content in flue gas is set as o. low The upper limit for residual oxygen content in flue gas is o. high In a preferred embodiment, o low Set to 1.8%, o high Set to 2.2%.
[0104] If the superheated steam flow rate is in a steady state at p future moments, then the following judgment is made:
[0105] If the residual oxygen content in the flue gas is below 0 low If the residual oxygen content in the flue gas continues to decrease within the set time, then the airflow rate of the blower and the airflow rate of the burner are increased by increasing the frequency of the blower and increasing the given opening degree of the burner air valve, until the residual oxygen content in the flue gas is not lower than 0. low ;
[0106] If the residual oxygen content in the flue gas is higher than o high If the residual oxygen content in the flue gas continues to rise within the set time, the blower's airflow and the burner's airflow will be reduced by decreasing the frequency of the blower and decreasing the given opening of the burner's air valve, until the residual oxygen content in the flue gas does not exceed 0. high ;
[0107] If the superheated steam flow rate is about to increase at future p time points, then first increase the gas flow rate and the air flow rate of the burner: This requires first increasing the given opening degree of the burner gas inlet valve, then increasing the given frequency of the blower's variable frequency drive, and finally increasing the given opening degree of the burner air valve; then the following judgments are made:
[0108] If the residual oxygen content in the flue gas is below 0 low If the residual oxygen content in the flue gas continues to decrease within the set time, then increase the air-to-coal ratio and increase the air flow rate of the blower. In this case, first increase the frequency of the blower's inverter, then increase the given opening degree of the burner air valve until the residual oxygen content in the flue gas is not lower than 0. low ;
[0109] If the residual oxygen content in the flue gas is higher than o high If the residual oxygen content in the flue gas continues to rise within the set time, reduce the air-coal ratio and reduce the air flow rate of the blower: at this time, the frequency of the blower should be reduced first, and then the air valve of the burner should be reduced to the given opening degree.
[0110] If the superheated steam flow rate is about to decrease at p future moments, then first reduce the gas flow rate and then the air flow rate of the burner: This requires first reducing the given opening of the burner gas inlet valve, then reducing the given frequency of the blower's frequency converter, and finally reducing the given opening of the burner air valve; then the following judgments are made:
[0111] If the residual oxygen content in the flue gas is below 0 low If the residual oxygen content in the flue gas continues to decrease within the set time, then increase the air-to-coal ratio and increase the air flow rate of the blower. In this case, first increase the frequency of the blower's inverter, then increase the given opening degree of the burner air valve until the residual oxygen content in the flue gas is not lower than 0. low ;
[0112] If the residual oxygen content in the flue gas is higher than o high If the residual oxygen content in the flue gas continues to rise within the set time, reduce the air-to-coal ratio and decrease the air flow rate of the blower. In this case, first reduce the frequency of the blower's inverter, then reduce the given opening of the burner air valve until the residual oxygen content in the flue gas is no higher than 0. high .
[0113] In a preferred embodiment, the set time for controlling the residual oxygen content in the flue gas is three minutes.
[0114] 3. Control circuit for negative pressure in the furnace.
[0115] 3.1 The control principle of furnace negative pressure:
[0116] To maintain normal boiler operation, air required for fuel combustion must be continuously supplied into the furnace, while the combustion products—flue gas—must be continuously discharged from the furnace. The process of supplying air into the furnace is called forced draft, and the process of discharging flue gas from the boiler to the outside is called induced draft. Insufficient negative pressure, or even positive pressure, in the furnace can cause flames to shoot out, damaging boiler equipment or causing burns; excessive negative pressure will draw in too much cold air (leaking in through leaks in the furnace, furnace doors, etc.), lowering the furnace temperature and increasing heat loss. The amount of forced draft depends on the combustion requirements. Whether the airflow is appropriate can be judged not only by using specialized instruments for analysis but also by observing the color of the furnace flame and flue gas. When the airflow is appropriate, the flame is wheat-yellow (bright yellow) and the flue gas is grayish-white; when the airflow is too high, the flame is bright white and the flue gas is white; when the airflow is too low, the flame is dark yellow or dark red and the flue gas is black.
[0117] When controlling the furnace negative pressure, the air supply flow rate is used as a feedforward variable. When the load is low, the coal feed rate is low, and the combustion intensity is very low, a low furnace pressure (absolute negative value) should be maintained to ensure effective improvement in thermal efficiency. Conversely, when the load is high, the coal feed rate is high, and the combustion intensity is very high, a high furnace pressure (absolute negative value) should be maintained to ensure effective improvement in boiler thermal efficiency. Adjusting the boiler furnace negative pressure mainly involves regulating the air supply and induced draft flow rates. A large air supply flow rate and a small induced draft flow rate result in a low furnace negative pressure (absolute negative value); conversely, a small air supply flow rate and a large induced draft flow rate result in a high furnace negative pressure (absolute negative value). Generally, a negative pressure of -150 Pa to -50 Pa should be maintained at the furnace outlet. When increasing the air volume, the induced draft flow rate should be increased first, followed by the air supply flow rate; when decreasing the air volume, the air supply flow rate should be decreased first, followed by the induced draft flow rate. In addition, the furnace negative pressure also affects the residual oxygen content in the flue gas.
[0118] 3.2 The specific control logic for furnace negative pressure is as follows: By adjusting the induced draft fan flow rate, the furnace negative pressure is controlled within the set range, specifically including:
[0119] The negative pressure data points in the furnace are generally distributed on the upper and lower layers of the front wall, the upper and lower layers of the rear wall, and the lower layer of the rear wall. Therefore, the negative pressure in the boiler furnace can be controlled by the ventilation volume of the upper and lower layers of the front wall, the upper and lower layers of the rear wall, respectively.
[0120] Let p be the set lower limit of the furnace negative pressure. low The upper limit of the negative pressure in the furnace is p high The unit for furnace negative pressure is Pa.
[0121] If the superheated steam flow rate is in a steady state at p future moments, then the following judgment is made:
[0122] When the boiler is running smoothly, the air flow rate of the blower is relatively stable. If the furnace negative pressure is higher than p... high If the supplied air flow is too large and the induced draft flow is too small, it is necessary to increase the frequency of the induced draft fan to increase the induced draft flow until the furnace negative pressure is no higher than p. high If the furnace negative pressure is lower than p low If the supplied air flow is too low and the induced draft flow is too high, it is necessary to reduce the induced draft fan's variable frequency to decrease the induced draft fan's induced draft flow until the furnace negative pressure is adjusted to the range required by the process, i.e., the furnace negative pressure is not lower than p. low ;
[0123] If the superheated steam flow rate is about to increase at future p time points, then first increase the gas flow rate and the air flow rate of the burner: This requires first increasing the given opening degree of the burner gas inlet valve, then increasing the given frequency of the blower's variable frequency drive, and finally increasing the given opening degree of the burner air valve; then the following judgments are made:
[0124] If the furnace negative pressure is higher than p high If the supplied air flow is too large and the induced draft flow is too small, it is necessary to increase the frequency of the induced draft fan to increase the induced draft flow until the furnace negative pressure is controlled and adjusted to the range required by the process, that is, the furnace negative pressure does not exceed p. higl ;
[0125] If the furnace negative pressure is lower than p iow If the supplied air flow is too low and the induced draft flow is too high, it is necessary to reduce the induced draft fan's variable frequency to decrease the induced draft fan's induced draft flow until the furnace negative pressure is adjusted to the range required by the process, i.e., the furnace negative pressure is not lower than p. low ;
[0126] If the superheated steam flow rate is about to decrease at p future moments, then first reduce the gas flow rate and then the air flow rate of the burner: This requires first reducing the given opening of the burner gas inlet valve, then reducing the given frequency of the blower's frequency converter, and finally reducing the given opening of the burner air valve; then the following judgments are made:
[0127] If the furnace negative pressure is higher than p high Then, the induced draft fan's airflow is increased by increasing its frequency converter until the furnace negative pressure is adjusted to the range required by the process, i.e., the furnace negative pressure does not exceed p. high ;
[0128] If the furnace negative pressure is lower than p low Then, the induced draft fan's airflow is reduced by decreasing its frequency converter until the furnace negative pressure is adjusted to the range required by the process, i.e., the furnace negative pressure is not lower than p. low .
[0129] 4. For the steam drum water level control loop.
[0130] 4.1 Control Principle of Steam Drum Water Level
[0131] Based on the current value of the superheated steam flow rate and its changing trend, combined with the current value of the steam drum water level, the main feedwater flow rate is adjusted to control the target value of the steam drum water level, keeping it within the range of -50mm to 50mm. The adjustment is achieved by using a PID control algorithm to adjust the valve opening of the main feedwater regulating valve and by fine-tuning the frequency converter frequency of the feedwater pump based on the pressure in the downstream header of the feedwater pump.
[0132] 4.2 The specific control logic for the steam drum water level is as follows: By adjusting the main feedwater flow rate, the steam drum water level is controlled within the set range, specifically including:
[0133] If the superheated steam flow rate is stable at p future moments, the main feedwater regulating valve is adjusted according to the real-time drum water level: the trend of the drum water level change is judged in advance; if the drum water level shows a downward trend, the opening of the main feedwater regulating valve is increased; if the drum water level shows an upward trend, the opening of the main feedwater regulating valve is decreased. The pressure difference between the feedwater pump downstream header pressure and the drum pressure is monitored in real time, and the drum water level and the pressure difference between the feedwater pump downstream header pressure and the drum pressure are controlled within the set range by adjusting the frequency of the feedwater pump.
[0134] If the superheated steam flow rate is about to increase at p future moments, the water level in the steam drum will drop and the steam drum pressure will increase. At this time, the main feedwater regulating valve of the boiler will be opened to increase the main feedwater flow rate in the steam drum and raise the water level in the steam drum. At the same time, due to the increase in the main feedwater flow rate, the pressure in the feedwater pump downstream header will decrease accordingly. In order to maintain the pressure difference between the feedwater pump downstream header pressure and the steam drum pressure within a process requirement range (in the preferred embodiment, the pressure difference needs to be controlled between 2 MPa and 2.5 MPa), the frequency of the feedwater pump needs to be increased by a given value to increase the feedwater pump flow rate and the feedwater pump downstream header pressure until the superheated steam flow rate reaches a stable state and the steam drum water level and the pressure difference between the feedwater pump downstream header pressure and the steam drum pressure are within the set range.
[0135] If the superheated steam flow rate is about to decrease at p future moments, the water level in the steam drum will rise and the steam drum pressure will decrease. At this time, the boiler main feedwater regulating valve will be reduced by a given opening to reduce the main feedwater flow rate in the steam drum and lower the steam drum water level. At the same time, due to the decrease in the boiler main feedwater flow rate, the pressure in the feedwater pump downstream header will increase accordingly. In order to maintain the pressure difference between the feedwater pump downstream header pressure and the steam drum pressure within a process requirement range, the frequency converter of the feedwater pump needs to be reduced by a given value to reduce the feedwater pump flow rate and the feedwater pump downstream header pressure until the superheated steam flow rate reaches a stable state and the steam drum water level and the pressure difference between the feedwater pump downstream header pressure and the steam drum pressure are within the set range.
[0136] 5. Control loop for superheated steam pressure.
[0137] 5.1 Control principle of superheated steam pressure:
[0138] Based on the current value of superheated steam flow and its changing trend, a model for the changing trends of superheated steam temperature and pressure, i.e., the characteristic curve of boiler load, is established using data such as superheated steam flow, superheated steam temperature, and superheated steam pressure. This provides preliminary operating condition information for the control of superheated steam temperature and pressure. The target pressure for superheated steam pressure is determined using the predicted superheated steam flow and the characteristic curve of boiler load.
[0139] The superheated steam pressure of a gas-fired boiler is mainly affected by the flow rate of the combustion gas, the boiler feedwater pump, and the main feedwater flow rate controlled by the main feedwater regulating valve. In addition, the changing trend of the superheated steam flow rate also has a significant impact on the real-time fluctuations of the superheated steam pressure. Therefore, a suitable coal-to-water ratio and the changing trend of the superheated steam flow rate can help the control system stably control the superheated steam pressure within a reasonable range.
[0140] 5.2 The specific control logic for superheated steam pressure is as follows: The superheated steam pressure is controlled within a set range by adjusting the gas flow rate of the burner, the air flow rate of the blower, and the frequency conversion frequency of the feedwater pump. Specifically, this includes:
[0141] If the superheated steam flow rate is in a steady state at p future moments, then there is no need to adjust the superheated steam pressure.
[0142] If the superheated steam flow rate is about to increase at p future moments, firstly, determine whether it is necessary to increase the main feedwater flow rate into the steam drum based on the current steam drum water level: Let the set upper limit of the steam drum water level be w. high The lower limit of the steam drum water level is w low If the current steam drum water level is higher than w high In this case, it is not necessary to adjust the main feedwater flow rate; if the current steam drum water level is higher than 0mm and lower than w high This will increase the main water supply flow rate by |ΔS t+p If the steam drum water level is below 0mm, first increase the main feedwater flow rate until the steam drum water level reaches 0mm, then increase the main feedwater flow rate by |ΔS t+p Secondly, monitor the pressure difference between the feedwater pump downstream main pipe pressure and the steam drum pressure, adjust the feedwater pump frequency converter frequency, and control the steam drum water level and the pressure difference between the feedwater pump downstream main pipe pressure and the steam drum pressure within the set range; finally, while maintaining the furnace negative pressure and the blast furnace gas pipeline pressure within the process requirements, increase the opening of the burner gas regulating valve until the superheated steam flow reaches the superheated steam flow predicted by the prediction model at a specific time. At the same time, increase the blower air flow to control the residual oxygen content in the flue gas within the set range.
[0143] If the superheated steam flow rate is about to decrease at p future moments, first determine whether it is necessary to reduce the main feedwater flow rate into the steam drum based on the current steam drum water level: Let w be the set upper limit of the steam drum water level. high The lower limit of the steam drum water level is w low If the current steam drum water level is higher than w high At this point, first adjust the steam drum water level to w. high Then reduce the main water supply flow rate by |ΔS t+p |;If the current steam drum water level is higher than 0mm and lower than w high This will reduce the main water supply flow rate by |ΔS t+p If the current steam drum water level is below 0mm, increase the main feedwater flow rate until the steam drum water level reaches 0mm, then increase the main feedwater flow rate by |ΔS t+p Secondly, monitor the pressure difference between the feedwater pump's downstream main pipe pressure and the steam drum pressure, adjust the feedwater pump's frequency converter frequency, and control the steam drum water level and the pressure difference between the feedwater pump's downstream main pipe pressure and the steam drum pressure within the set range. Finally, while maintaining the furnace negative pressure and the blast furnace gas pipeline pressure within the process requirements, reduce the opening of the burner's gas regulating valve until the superheated steam flow rate reaches the superheated steam flow rate predicted by the prediction model at a specific time. At the same time, reduce the blower's air flow rate to control the residual oxygen content in the flue gas within the set range.
[0144] 6. Establish a steam temperature control model based on energy balance for the control loops of superheated steam temperature and reheated steam temperature.
[0145] 6.1 Control principles for superheated steam temperature and reheated steam temperature:
[0146] The changes in superheated steam temperature and reheated steam temperature are related to the changes in the superheated steam flow rate of the boiler. The increase in superheated steam temperature is caused by the increase in boiler combustion load. The increase in combustion load will lead to an increase in flue gas temperature in the furnace and flue, which in turn will lead to an increase in the wall temperature of the superheater and reheater, ultimately resulting in an increase in superheated steam temperature and reheated steam temperature.
[0147] 6.2 The specific control logic for superheated steam temperature is as follows: The superheated steam temperature is controlled within the set range by adjusting the flow rate of the desuperheating water in the secondary superheated steam desuperheater, specifically including:
[0148] Using the superheated steam temperature after the outlet header of the screen-type superheater as feedforward information and the superheated steam temperature after the outlet header of the high-temperature superheater as the control target, the superheated steam temperature is controlled by adjusting the opening of the regulating valve of the superheated steam secondary desuperheater.
[0149] According to the principle of energy balance, the working fluid energy is balanced before and after the superheated steam desuperheater after the outlet header of the screen-type superheater. The working fluid includes steam and water. Therefore, the energy of the steam pipe before the desuperheating water is:
[0150] L s ×H s (P s ,T s )+L w ×H w (P w ,T w );
[0151] Among them, L s H s P s T s These are the superheated steam flow rate, superheated steam enthalpy, superheated steam pressure, and superheated steam temperature before the secondary superheated steam desuperheater; L w H w P w T w These are the desuperheating water flow rate, desuperheating water enthalpy, desuperheating water pressure, and desuperheating water temperature of the secondary desuperheating unit for superheated steam.
[0152] The energy of the steam pipeline after the superheated steam secondary desuperheater is L. out ×H out (P out ,T out ); where L out H out P out T out These are the superheated steam flow rate, superheated steam enthalpy, superheated steam pressure, and superheated steam temperature after the second-stage superheated steam desuperheater;
[0153] Energy balance formula:
[0154] L s ×H s (P s ,T s )+L w ×H w (P w ,T w ) = L out ×H out (P out ,T out );
[0155] Among them, the desuperheating water flow rate L of the superheated steam secondary desuperheater w The unknown quantity is the temperature T of the superheated steam after the secondary desuperheater. outThe superheated steam temperature is an indirect target temperature. There is a relatively definite correlation between the superheated steam temperature and the superheated steam flow rate. The control target temperature of the superheated steam temperature is determined by combining the predicted superheated steam flow rate with the boiler load characteristic curve. The boiler load characteristic curve is obtained by analyzing historical data.
[0156] The desuperheating water flow rate L of the secondary desuperheater of superheated steam is derived using the energy balance formula. w :
[0157]
[0158] L w The value is used as the target value, and the flow rate of the desuperheating water is adjusted by the superheater desuperheating water regulating valve to control the superheated steam temperature within the set range.
[0159] 6.3 The specific control logic for reheat steam temperature is as follows: Controlling the reheat steam temperature within the set range by adjusting the reheat steam desuperheating water flow rate and flue gas flow rate specifically includes:
[0160] There is a correlation between reheat steam temperature and superheated steam flow rate. The target temperature for reheat steam temperature is determined by combining the predicted superheated steam flow rate with the boiler load characteristic curve.
[0161] By monitoring the reheat steam temperature after the high-temperature reheater and the rate of change of the reheat steam temperature after the high-temperature reheater, the opening of the flue gas electric regulating damper valve is adjusted using a PID algorithm to control the reheat steam temperature within the set range.
[0162] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention, and no reference numerals in the claims should be construed as limiting the scope of the claims.
[0163] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.
Claims
1. A method for controlling gas boilers in power plants based on prediction of machine learning models. It is characterized in that The following steps are involved: Step 1: Establish a prediction model for superheated steam flow based on a machine learning model: A production data set D1 consisting of historical production data of gas boilers in a power plant is obtained; the production data includes the following variables: blast furnace gas network flow, blast furnace gas network pressure, furnace negative pressure, blower air flow, flue gas residual oxygen content, blast furnace gas calorific value, air-coal ratio, turbine generator set power generation and superheated steam flow, wherein the superheated steam flow is the target variable, and the remaining variables are prediction variables, which are used to train the prediction model; first, the production data set D1 is clustered based on the variables, and the production data set D1 is divided into N types of training samples, and a prediction model is trained by each type of training sample, and a total of N prediction models are obtained; in actual production, the current production data is first classified to obtain the classification result; according to the classification result, the current production data is input into the corresponding trained prediction model, and the prediction model predicts the superheated steam flow at p moments in the future; Step 2: Determine the state of the superheated steam flow rate at the next p moments: The superheated steam flow rate at the current moment t is S t , the standard deviation of the superheated steam flow rate in the past p moments is σ t-p The superheated steam flow rate predicted by the prediction model at the next p moments is S t+1 , S t+2 , …, S t+p , the change in superheated steam flow rate after p moments ΔS t+p =S t+p -S t ; If |ΔS t+p |≤|σ t-p |, the superheated steam flow rate at the next p moments is considered to be in a stable state; If |ΔS t+p |>|σ t-p |, the superheated steam flow rate in the next p moments is considered to be in an unsteady state; and when ΔS t+p >0,σ t-p >0, the superheated steam flow rate at the next p moments will increase; and when ΔS t+p <0,σ t-p When <0, the superheated steam flow rate at the next p moments will decrease; Step three, combining the state of superheated steam flow at p moments in the future, control the flue gas residual oxygen content, furnace negative pressure, drum water level, superheated steam pressure, superheated steam temperature and reheated steam temperature at the current moment.
2. The power plant gas boiler control method based on machine learning model prediction according to claim 1, It is characterized in that In step 3, the control of flue gas residual oxygen content, furnace negative pressure, drum water level, superheated steam pressure, superheated steam temperature and reheated steam temperature specifically includes: By adjusting the gas flow rate of the burner and the air flow rate of the burner, the residual oxygen content of the flue gas can be controlled within the set range; By adjusting the induced draft fan flow rate, the negative pressure in the furnace can be controlled within the set range; By adjusting the main feed water flow, the drum water level is controlled within the set range; The superheated steam pressure is controlled within the set range by adjusting the gas flow of the burner, the air flow of the blower and the variable frequency of the water supply pump; The superheated steam temperature is controlled within the set range by adjusting the desuperheating water flow of the superheated steam secondary desuperheater; The reheat steam temperature is controlled within the set range by adjusting the reheat steam cooling water flow and flue gas flow.
3. The power plant gas boiler control method based on machine learning model prediction according to claim 2, It is characterized in that The method of controlling the residual oxygen content of the flue gas within a set range by adjusting the gas flow rate of the burner and the air flow rate of the burner specifically includes: Note that the lower limit of residual oxygen content in flue gas is set as o low , the upper limit of flue gas residual oxygen content is o high ; If the superheated steam flow rate is stable at the next p moments, the following judgment is made: If the residual oxygen content of flue gas is lower than o low If the residual oxygen content of the flue gas is still decreasing within the set time, the fan frequency is increased and the burner air valve is opened to a given degree to increase the air flow of the fan and the burner air flow until the residual oxygen content of the flue gas is not less than o low ; If the residual oxygen content of flue gas is higher than o high If the residual oxygen content of the flue gas is still rising within the set time, the fan frequency and the burner air valve opening are reduced to a given degree to reduce the fan air flow and the burner air flow until the residual oxygen content of the flue gas is no higher than o high ; If the superheated steam flow rate is about to increase at p moments in the future, the gas flow rate of the burner is increased first, and then the air flow rate of the burner is increased: at this time, the burner gas inlet valve needs to be increased by a given opening, and then the variable frequency of the blower is increased by a given frequency, and finally the burner air valve is increased by a given opening; then the following judgment is made: If the residual oxygen content of flue gas is lower than o low If the residual oxygen content in the flue gas is still decreasing within the set time, increase the air-to-coal ratio and increase the air flow rate of the blower: at this time, increase the frequency of the blower first, and then increase the burner air valve to a given opening until the residual oxygen content in the flue gas is not less than o low ; If the residual oxygen content of flue gas is higher than o high , and the residual oxygen content of the flue gas is still rising within the set time, then reduce the air-to-coal ratio and reduce the air flow rate of the blower: at this time, first reduce the frequency of the blower, and then reduce the burner air valve to a given opening; If the superheated steam flow rate is about to decrease at the next p moments, the gas flow rate of the burner is reduced first, and then the air flow rate of the burner is reduced: at this time, the burner gas inlet valve needs to be reduced by a given opening, and then the frequency conversion of the blower needs to be reduced by a given frequency, and finally the burner air valve needs to be reduced by a given opening; then the following judgment is made: If the residual oxygen content of flue gas is lower than o low If the residual oxygen content in the flue gas is still decreasing within the set time, increase the air-to-coal ratio and increase the air flow rate of the blower: at this time, increase the frequency of the blower first, and then increase the burner air valve to a given opening until the residual oxygen content in the flue gas is not less than o low ; If the residual oxygen content of flue gas is higher than o high If the residual oxygen content in the flue gas is still rising within the set time, reduce the air-to-coal ratio and reduce the air flow rate of the blower: at this time, first reduce the frequency of the blower, and then reduce the burner air valve to a given opening until the residual oxygen content in the flue gas is no higher than o high .
4. The power plant gas boiler control method based on machine learning model prediction according to claim 2, It is characterized in that The method of controlling the furnace negative pressure within a set range by adjusting the induced draft fan induced air flow rate specifically includes: The lower limit of the furnace negative pressure is set as p low , the upper limit of furnace negative pressure is p high ; If the superheated steam flow rate is stable at the next p moments, the following judgment is made: If the furnace negative pressure is higher than p high , then the induced draft fan flow rate is increased by increasing the frequency conversion of the induced draft fan until the negative pressure in the furnace is no higher than p high ; If the furnace negative pressure is lower than p low , then reduce the induced draft fan flow rate by reducing the frequency conversion of the induced draft fan until the negative pressure in the furnace is not less than p low ; If the superheated steam flow rate is about to increase at p moments in the future, the gas flow rate of the burner is increased first, and then the air flow rate of the burner is increased: at this time, the burner gas inlet valve needs to be increased by a given opening, and then the variable frequency of the blower is increased by a given frequency, and finally the burner air valve is increased by a given opening; then the following judgment is made: If the furnace negative pressure is higher than p high , then the induced draft fan flow rate is increased by increasing the frequency conversion of the induced draft fan until the negative pressure in the furnace is no higher than p high ; If the furnace negative pressure is lower than p low , then reduce the induced draft fan flow rate by reducing the frequency conversion of the induced draft fan until the negative pressure in the furnace is not less than p low ; If the superheated steam flow rate is about to decrease at the next p moments, the gas flow rate of the burner is reduced first, and then the air flow rate of the burner is reduced: at this time, the burner gas inlet valve needs to be reduced by a given opening, and then the frequency conversion of the blower needs to be reduced by a given frequency, and finally the burner air valve needs to be reduced by a given opening; then the following judgment is made: If the furnace negative pressure is higher than p high , then the induced draft fan flow rate is increased by increasing the frequency conversion of the induced draft fan until the negative pressure in the furnace is no higher than p high ; If the furnace negative pressure is lower than p low , then reduce the induced draft fan flow rate by reducing the frequency conversion of the induced draft fan until the negative pressure in the furnace is not less than p low .
5. The power plant gas boiler control method based on machine learning model prediction according to claim 2, It is characterized in that The method of controlling the drum water level within a set range by adjusting the main feed water flow rate specifically includes: If the superheated steam flow rate at p moments in the future is in a stable state, the main feedwater regulating valve is adjusted according to the real-time drum water level: the changing trend of the drum water level is judged. If the drum water level has a downward trend, the main feedwater regulating valve opening is increased; if the drum water level has an upward trend, the main feedwater regulating valve opening is reduced, and the pressure difference between the main pipe pressure after the feedwater pump and the drum pressure is monitored. By adjusting the frequency conversion frequency of the feedwater pump, the pressure difference between the drum water level, the main pipe pressure after the feedwater pump and the drum pressure is controlled within the set range; If the superheated steam flow rate is about to increase at p moments in the future, the water level in the drum will drop and the drum pressure will increase. At this time, the boiler main feed water regulating valve will be increased by a given opening to increase the main feed water flow rate of the drum and raise the drum water level; the variable frequency of the feed water pump will be increased by a given value to increase the feed water flow rate of the feed water pump and the pressure of the main pipe behind the feed water pump until the superheated steam flow rate reaches a stable state and the pressure difference between the drum water level and the pressure of the main pipe behind the feed water pump and the drum pressure is within the set range; If the superheated steam flow rate is about to decrease at p moments in the future, the water level in the drum will rise and the drum pressure will decrease. At this time, the boiler main feed water regulating valve will be reduced by a given opening to reduce the main feed water flow rate of the drum and lower the drum water level; the variable frequency of the feed water pump will be reduced by a given value to reduce the feed water flow rate of the feed water pump and the main pipe pressure behind the feed water pump, until the superheated steam flow rate reaches a stable state, and the pressure difference between the drum water level and the main pipe pressure behind the feed water pump and the drum pressure is within the set range.
6. The power plant gas boiler control method based on machine learning model prediction according to claim 2, It is characterized in that The superheated steam pressure is controlled within a set range by adjusting the gas flow of the burner, the air flow of the blower and the variable frequency of the water supply pump, specifically including: If the superheated steam flow rate at the next p moments is stable, there is no need to adjust the superheated steam pressure; If the superheated steam flow rate is about to increase at the next p moments, first, determine whether it is necessary to increase the main feed water flow entering the drum based on the current drum water level: the upper limit of the drum water level is set as w high , the lower limit of the drum water level is w low ; If the current drum water level is higher than w high , then there is no need to adjust the main feed water flow rate; if the current drum water level is higher than 0mm and lower than w high , then increase the main water flow rate by |ΔS t+p |; If the drum water level is lower than 0mm, first increase the main feed water flow until the drum water level reaches 0mm, then increase the main feed water flow by |ΔS t+p |; Secondly, monitor the pressure difference between the main pipe pressure behind the feedwater pump and the drum pressure, adjust the frequency conversion frequency of the feedwater pump, and control the drum water level and the pressure difference between the main pipe pressure behind the feedwater pump and the drum pressure within the set range; finally, increase the opening of the burner gas regulating valve under the premise of maintaining the furnace negative pressure and the blast furnace gas network pressure within the process requirements, until the superheated steam flow at a specific time reaches the superheated steam flow at the corresponding time predicted by the prediction model. At the same time, increase the air flow of the blower to control the residual oxygen content of the flue gas within the set range; If the superheated steam flow rate is about to decrease at the next p moments, first determine whether it is necessary to reduce the main feed water flow entering the drum based on the current drum water level: the upper limit of the drum water level is set as w high , the lower limit of the drum water level is w low ; If the current drum water level is higher than w high , then at this time, adjust the drum water level to w high , and then reduce the main water flow rate by |ΔS t+p |; If the current drum water level is higher than 0mm and lower than w high , then the main feed water flow rate is reduced by |ΔS t+p |; If the current drum water level is lower than 0mm, increase the main feed water flow until the drum water level reaches 0mm, and then increase the main feed water flow by |ΔS t+p |; Secondly, monitor the pressure difference between the main pipe pressure behind the feedwater pump and the drum pressure, adjust the frequency conversion frequency of the feedwater pump, and control the drum water level and the pressure difference between the main pipe pressure behind the feedwater pump and the drum pressure within the set range; finally, reduce the opening of the burner's gas regulating valve while maintaining the furnace negative pressure and the blast furnace gas network pressure within the process requirements, until the superheated steam flow at a specific moment reaches the superheated steam flow at the corresponding moment predicted by the prediction model. At the same time, reduce the air flow of the blower to control the residual oxygen content in the flue gas within the set range.
7. The power plant gas boiler control method based on machine learning model prediction according to claim 2, It is characterized in that The method of controlling the temperature of superheated steam within a set range by adjusting the desuperheating water flow rate of the secondary desuperheater of superheated steam specifically includes: The superheated steam temperature after the platen superheater outlet header is used as the feedforward information, and the superheated steam temperature after the high-temperature superheater outlet header is used as the control target. The superheated steam temperature is controlled by adjusting the opening of the regulating valve of the superheated steam secondary desuperheater cooling water: According to the energy balance principle, the energy balance of the working medium before and after the superheated steam desuperheater after the platen superheater outlet header is that the working medium includes steam and water. The energy of the steam pipe before the desuperheating water is injected is L s ×H s (P s , T s )+L w ×H w (P w , T w ), where L s , H s , P s , T s are respectively the superheated steam flow rate, superheated steam enthalpy, superheated steam pressure and superheated steam temperature before the superheated steam secondary desuperheater; L w , H w , P w , T w They are respectively the desuperheating water flow rate, desuperheating water enthalpy value, desuperheating water pressure and desuperheating water temperature of the secondary desuperheater of superheated steam; The energy of the steam pipe after the secondary desuperheater of superheated steam is L out ×H out (P out , T out ), where L out , H out , P out , T out They are the superheated steam flow rate, superheated steam enthalpy, superheated steam pressure and superheated steam temperature after the superheated steam secondary desuperheater; Energy balance formula: L s ×H s (P s ,T s )+L w ×H w (P w ,T w )=L out ×H out (P out ,T out ); Among them, the cooling water flow rate L of the superheated steam secondary desuperheater w is an unknown quantity, the superheated steam temperature after the secondary desuperheater is T out It is the indirect target temperature of superheated steam temperature. There is a correlation between superheated steam temperature and superheated steam flow rate. The control target temperature of superheated steam temperature is determined by combining the predicted superheated steam flow rate with the boiler load characteristic curve. The cooling water flow rate L of the superheated steam secondary desuperheater is derived by the energy balance formula: w : L w The value is used as the target value, and the superheater desuperheating water flow rate is adjusted through the superheater desuperheating water regulating valve to control the superheated steam temperature within the set range.
8. The power plant gas boiler control method based on machine learning model prediction according to claim 2, It is characterized in that The method of controlling the reheat steam temperature within a set range by adjusting the reheat steam desuperheating water flow rate and the flue gas flow rate specifically includes: There is a relatively certain correlation between the reheat steam temperature and the superheated steam flow rate. The control target temperature of the reheat steam temperature is determined by combining the predicted superheated steam flow rate with the boiler load characteristic curve. By monitoring the reheat steam temperature after the high-temperature reheater and the temperature change rate of the reheat steam after the high-temperature reheater, the PID algorithm is used to adjust the opening of the flue gas electric regulating damper valve to control the reheat steam temperature within the set range.
9. The power plant gas boiler control method based on machine learning model prediction according to claim 1, It is characterized in that The prediction model adopts the LightGBM regression model.
Citation Information
Patent Citations
Method for predicting steam yield of steam boiler based on data mining
CN112906307A
Steel rolling heating furnace control method based on machine learning model
CN115186582A
Automatic combustion control system for three kinds of coal gas of coal gas power generation boiler
CN115681943A
Control system of gas-fired boiler
CN115751276A
Power plant boiler operation control method based on XGBoost optimization
CN116300666A
Cited By
Fan frequency control method and device, electronic equipment and computer storage medium
CN121206497A
Fuel oil exhaust gas waste heat power generation method and device, electronic equipment and readable storage medium
CN122190861A
Fuel exhaust waste heat power generation method and device, electronic equipment, readable storage medium
CN122190861B
Self-adaptive control method for gas power generation load of power plant under condition of no gas holder or insufficient holder capacity
CN122246781A
