Unit heat correction method and system based on main steam pressure prediction

Abstract

The invention relates to the technical field of turbine-boiler coordination control, in particular to a unit heat correction method and system based on main steam pressure prediction, and the method comprises the steps: obtaining unit parameters; constructing a deep learning prediction model, inputting the standardized data into the deep learning prediction model, and obtaining a main steam pressure prediction value in a future prediction duration; calculating an energy balance signal and an original heat signal, and dynamically correcting the original heat signal according to the main steam pressure predicted value to obtain a corrected heat signal; and the corrected heat signal serves as a feedback value, the energy balance signal serves as a set value, the fuel flow and the feed water flow of the boiler are regulated and controlled, unit heat correction is achieved, the problems of passive adjustment and insufficient precision of a DEB control technology in the prior art are solved, and the purposes of stabilizing the main steam pressure of the unit and improving the energy efficiency are achieved.

Description

Technical Field

[0001] This invention relates to the field of boiler-turbine coordinated control technology, and in particular to a method and system for unit heat correction based on main steam pressure prediction. Background Technology

[0002] With the continuous increase in the proportion of new energy installed capacity, the role of thermal power has transformed from traditional base load power source to peak shaving and frequency regulation power source. Variable load operation of units has become the norm. At the same time, the dual carbon targets have put forward higher requirements for energy conservation and consumption reduction of thermal power units. The popularization of high-parameter units such as supercritical and ultra-supercritical units has further aggravated the difficulty of controlling the main steam pressure and heat balance.

[0003] Direct energy balance (DEB) control technology is a classic solution for coordinated control of the turbine and boiler in current thermal power units. Its core involves real-time acquisition of parameters such as main steam pressure, regulating stage pressure, and drum pressure to calculate the energy balance signal (BD), representing the turbine's energy demand, and the raw heat signal (HR), representing the boiler's heat production capacity. Then, PID control is used to ensure the boiler output tracks the turbine's demand, thereby maintaining stable main steam pressure. However, traditional DEB control technology has key limitations, becoming a core issue restricting the stability and energy efficiency of unit operation. First, the lack of a main steam pressure prediction mechanism results in a passive and lagging regulation mode: Traditional DEB relies entirely on real-time pressure feedback for regulation, without considering the future trend of main steam pressure changes. Under dynamic operating conditions such as load changes and start-up / shutdown, the main steam pressure deviation often reaches ±0.8MPa or more, making it unable to respond quickly to AGC load adjustment commands. This can easily lead to an imbalance between boiler and turbine energy supply and demand, affecting the stable operation of the unit. Secondly, the heat correction accuracy is insufficient and the adaptability is poor: the heat signal calculation of traditional DEB is based on a static model. When faced with dynamic changes such as coal quality fluctuations and turbine efficiency decline, the static error of heat correction reaches ±1%~2%, which directly leads to high coal consumption and makes it difficult to meet the energy-saving requirements of low-carbon transformation. Finally, the main steam pressure and heat balance of high-parameter units are strongly nonlinearly coupled, and multiple factors such as boiler combustion efficiency and ambient temperature interfere with each other. Traditional linear models are unable to accurately describe this relationship, which further exacerbates the fluctuation of main steam pressure and the deviation of heat correction.

[0004] Therefore, a heat correction method is needed that can predict the trend of main steam pressure changes, dynamically correct heat signals, and adapt to complex operating conditions and high-parameter units, in order to solve the problems of passive adjustment and insufficient accuracy of traditional DEB control technology, and achieve the goal of stabilizing main steam pressure and improving energy efficiency. Summary of the Invention

[0005] This application provides a unit heat correction method and system based on main steam pressure prediction, which solves the problems of passive adjustment and insufficient accuracy of DEB control technology in the prior art, and achieves the goal of stabilizing the main steam pressure and improving energy efficiency of the unit.

[0006] In a first aspect, this application provides a unit heat correction method based on main steam pressure prediction, including: S1. Obtain unit parameters, including main steam pressure, main steam pressure setpoint, regulating stage pressure, steam drum pressure, unit load command, fuel flow rate, and feedwater flow rate; S2. The unit parameters are preprocessed to obtain standardized data; S3. Construct a deep learning prediction model, input the standardized data into the deep learning prediction model, and obtain the predicted value of the main steam pressure within the future prediction period; S4. Calculate the energy balance signal based on the main steam pressure, the main steam pressure setpoint, and the regulating stage pressure; calculate the original heat signal based on the regulating stage pressure and the steam drum pressure. The energy balance signal is used to represent the turbine energy demand, and the original heat signal is used to represent the boiler heat production capacity. Dynamically correct the original heat signal according to the predicted main steam pressure to obtain the corrected heat signal. S5. Using the corrected heat signal as a feedback value and the energy balance signal as a set value, the boiler's fuel flow rate and feedwater flow rate are regulated to achieve unit heat correction.

[0007] Furthermore, the main steam pressure is the turbine inlet pressure, the regulating stage pressure is the turbine first stage pressure, the steam drum pressure is the pressure inside the steam drum of the steam drum boiler, the main steam pressure setpoint is generated according to the unit load command, and the unit load command is issued by the power grid AGC system.

[0008] Furthermore, the preprocessing of the unit parameters to obtain standardized data includes filtering the unit parameters using an adaptive filtering algorithm and mapping the filtered unit parameters to a preset numerical range using a normalization method.

[0009] Furthermore, the deep learning prediction model is a long short-term memory network model, which includes an input layer, an embedding layer, three LSTM layers, two fully connected layers, and an output layer. The number of hidden units in the three LSTM layers are 64, 32, and 16, respectively. The activation function of the LSTM layers is the tanh function, and the output layer uses a linear activation function.

[0010] Furthermore, the standardized data is input into a deep learning prediction model to obtain the predicted main steam pressure value within the future prediction period, including: S31. Construct an input feature sequence from the standardized data in chronological order. The input feature sequence includes time-series variation data of main steam pressure, main steam pressure setpoint, regulating stage pressure, steam drum pressure, unit load command, fuel flow rate, and feedwater flow rate. S32. The input feature sequence is mapped to a preset feature dimension through an embedding layer, and the mapped features are deeply extracted and time-dependent modeled through a 3-layer LSTM layer to obtain the nonlinear coupling relationship between the standardized data in the input feature sequence and the future trend of the main steam pressure. S33. The output features of the LSTM layer are compressed by gradient mapping through the two fully connected layers of the model, and then linearly mapped and transformed through the output layer to output the predicted value of the main steam pressure within the preset prediction time.

[0011] Furthermore, the calculation of the energy balance signal based on the main steam pressure, the main steam pressure setpoint, and the regulating stage pressure, and the calculation of the original heat signal based on the regulating stage pressure and the steam drum pressure, include: S41. The formula for calculating the energy balance signal is: ,in, , The preset adjustment coefficients for basic unit operation. Main steam pressure, Main steam pressure setpoint For regulating stage pressure, This is an energy balance signal.

[0012] S42. The formula for calculating the original heat signal is: ,in, Used to indicate changes in the energy storage of a boiler. For the steam drum pressure, This is the original heat signal.

[0013] Further, the step of dynamically correcting the original heat signal to obtain the corrected heat signal includes: S43. Based on the predicted main steam pressure value and the set main steam pressure value, calculate the prediction deviation rate between the two. The calculation formula is: ,in, This is the predicted value of steam pressure. The predicted deviation rate is used to quantify the deviation trend of future main steam pressure. S44. Set a preset correction coefficient, which is used to quantify the degree of influence of the prediction deviation on the original heat signal; S45. Based on the prediction deviation rate and the preset correction coefficient, the original heat signal is dynamically corrected, and the calculation formula is: ,in, This is a preset correction factor. This is the corrected heat signal.

[0014] Furthermore, the preset correction coefficient needs to be calibrated, and calibrating the preset correction coefficient includes: S441. Divide the full load range of the unit into multiple consecutive preset load ranges, set a corresponding preset coefficient range for each preset load range, and select multiple candidate coefficient values ​​at equal intervals based on the preset coefficient range; S442. Within each preset load range, the unit is stably operated at the intermediate load point of that range, and each of the candidate coefficient values ​​is sequentially used as a preset correction coefficient. Substituting the temporary values ​​into the dynamic correction formula Keep other control parameters constant, and calculate the fluctuation range of the main steam pressure based on the acquired main steam pressure data; S443. Set a preset fluctuation threshold, wherein the preset fluctuation threshold is the maximum main steam pressure fluctuation range allowed for safe and stable operation of the unit. Compare the fluctuation ranges corresponding to all candidate coefficient values ​​in each preset load interval, select the candidate coefficient value that satisfies the condition that the fluctuation range is less than the preset fluctuation threshold and the fluctuation range is the smallest, and determine it as the preset correction coefficient corresponding to the preset load interval. S444. Based on the current actual load of the unit, a linear interpolation algorithm is used to achieve smooth switching of the preset correction coefficient between adjacent preset load intervals.

[0015] Furthermore, the step of using the corrected heat signal as a feedback value and the energy balance signal as a set value to regulate the boiler's fuel flow rate and feedwater flow rate includes: S51. An adaptive PID control algorithm is adopted, with the corrected heat signal as the feedback value and the energy balance signal as the set value. Based on the deviation between the corrected heat signal and the energy balance signal, the proportional coefficient and integral coefficient are dynamically adjusted to adapt the control action to the changes in operating conditions. S52. Based on the adjusted adaptive PID regulation algorithm, output control signals to regulate the boiler's fuel flow and feedwater flow, so that the boiler output and the turbine's energy demand are dynamically matched. S53. The turbine main control unit receives the unit load command and the main steam pressure prediction value. When the deviation between the main steam pressure prediction value and the main steam pressure set value exceeds the preset deviation threshold, the turbine valve opening is adjusted to help stabilize the main steam pressure.

[0016] Furthermore, the regulation of the boiler's fuel flow rate and feedwater flow rate includes: S521. Obtain the optimal ratio data of fuel flow and feedwater flow under different load ranges of the current unit, and establish a load ratio mapping relationship table; S522. Based on the current actual load of the unit and the predicted main steam pressure, the corresponding optimal ratio is matched from the mapping table using an interpolation algorithm; S523. During the control process, the optimal ratio is used as a benchmark, and the ratio deviation is dynamically fine-tuned according to the real-time changes in the corrected heat signal.

[0017] Secondly, a unit heat correction system based on main steam pressure prediction includes: The data acquisition module is configured to acquire unit parameters, including main steam pressure, main steam pressure setpoint, regulating stage pressure, steam drum pressure, unit load command, fuel flow rate, and feedwater flow rate. The data preprocessing module is configured to preprocess the unit parameters to obtain standardized data; The main steam pressure prediction module is configured to construct a deep learning prediction model, input the standardized data into the deep learning prediction model, and obtain the predicted value of the main steam pressure within the future prediction period. The heat signal correction module is configured to calculate an energy balance signal based on the main steam pressure, the main steam pressure setpoint, and the regulating stage pressure; calculate an original heat signal based on the regulating stage pressure and the steam drum pressure; the energy balance signal is used to represent the turbine energy demand; the original heat signal is used to represent the boiler heat production capacity; and dynamically correct the original heat signal according to the predicted main steam pressure to obtain a corrected heat signal. The coordination control module is configured to use the corrected heat signal as a feedback value and the energy balance signal as a set value to regulate the boiler's fuel flow and feedwater flow, thereby achieving unit heat correction.

[0018] Thirdly, the present invention provides a computer-readable storage medium storing a plurality of instructions adapted for loading and execution by a processor of a terminal device of the unit heat correction method based on main steam pressure prediction.

[0019] Fourthly, the present invention provides a terminal device, including a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, the instructions being adapted to be loaded and executed by the processor for the unit variable rate control method based on main gas pressure prediction.

[0020] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: 1. This application constructs a long short-term memory network model to predict main steam pressure and dynamically corrects the original heat signal based on the predicted value, effectively solving the passive adjustment problem caused by the lack of pressure prediction and reliance on real-time feedback in traditional DEB control technology. Specifically, by inputting the time series data of the unit's full parameters, the model captures the nonlinear coupling relationship between core parameters such as fuel flow and feedwater flow and the future trend of main steam pressure. Combined with the prediction deviation rate, the heat signal is pre-corrected, enabling the boiler output to adapt to the turbine's energy demand in advance. This significantly reduces the main steam pressure fluctuation under dynamic operating conditions such as load changes and start-up / shutdown. The main steam pressure deviation can be controlled within a preset accuracy range, improving the unit's response speed and operational stability to AGC load adjustments.

[0021] 2. This application addresses the problem of poor adaptability of traditional static models to coal quality fluctuations and load changes by calibrating preset correction coefficients across load intervals and employing linear interpolation for smooth switching. Simultaneously, it utilizes adaptive filtering and normalization preprocessing to optimize input data quality. The interval calibration of the correction coefficients ensures targeted heat correction under different load conditions, while interpolation switching guarantees the continuity of the correction signal, achieving refined control of heat distribution and facilitating reduced coal consumption by the unit.

[0022] 3. This application employs an adaptive PID control algorithm to dynamically adjust the proportional and integral coefficients, combined with optimal fuel flow and feedwater flow ratio control and turbine fine-tuning, thus solving the energy supply and demand imbalance problem caused by poor adaptability and insufficient boiler-turbine coordination in traditional control algorithms. The adaptive PID algorithm can dynamically optimize the control parameters based on the deviation between the heat signal and the energy balance signal. Optimal ratio control ensures stable boiler combustion efficiency, and early turbine fine-tuning further assists in stabilizing the main steam pressure, enabling dynamic matching between boiler output and turbine energy demand. It is particularly suitable for the complex coupling characteristics of high-parameter units such as supercritical and ultra-supercritical units, ensuring stable and efficient operation of the unit within the 30%~100% rated load range. Attached Figure Description

[0023] Figure 1 This is a flowchart of the method in Embodiment 1 of this application; Figure 2 This is a historical trend graph of different pressure measuring points in Embodiment 1 of this application; Figure 3 This is a module diagram of Embodiment 2 of this application. Detailed Implementation

[0024] With the increasing proportion of new energy installed capacity and the advancement of dual-carbon goals, thermal power units need to undertake peak shaving and frequency regulation tasks. Variable load operation at 30% to 100% of rated load has become the norm. The widespread adoption of high-parameter units such as supercritical and ultra-supercritical units has further exacerbated the difficulty of controlling the main steam pressure and heat balance. Traditional direct energy balance (DEB) control technology relies on real-time pressure feedback regulation and lacks a main steam pressure prediction link, resulting in main steam pressure deviations often exceeding ±0.8 MPa under dynamic operating conditions. Furthermore, static heat correction errors cannot meet energy-saving requirements and are difficult to adapt to dynamic changes such as coal quality fluctuations and unit efficiency decline. This embodiment addresses this core problem by proposing a unit heat correction scheme that integrates deep learning prediction and dynamic heat correction. By predicting the main steam pressure change trend and accurately correcting the heat signal, it achieves active matching of boiler and turbine energy supply and demand, overcoming the shortcomings of passive regulation and insufficient accuracy in traditional technologies.

[0025] The core technical solution of this embodiment is achieved through five collaborative steps: First, seven key unit parameters, including main steam pressure, regulating stage pressure, and drum pressure, are acquired and standardized data is obtained through adaptive filtering and normalization preprocessing. Then, a Long Short-Term Memory (LSTM) network model is constructed, inputting the time-series feature sequence of the standardized data to accurately predict changes in main steam pressure within a preset timeframe. Based on measured parameters, the energy balance signal (BD), representing the turbine's energy demand, and the raw heat signal (HR), representing the boiler's heat production capacity, are calculated. The deviation rate is calculated using the predicted main steam pressure value, and the HR is dynamically corrected using correction coefficients calibrated for load distribution. Finally, using the corrected heat signal as feedback and the energy balance signal as setpoint, an adaptive PID algorithm is used to regulate boiler fuel flow and feedwater flow, coordinating with turbine valve fine-tuning to form a closed-loop control. This solution deeply integrates data-driven prediction with the traditional DEB mechanism, achieving an improvement in the regulation mode from passive feedback to active prediction.

[0026] The key technologies in this embodiment are: First, the structured construction of the LSTM prediction model, which captures the nonlinear coupling relationship between parameters such as fuel flow rate and feedwater flow rate and main steam pressure to ensure that the prediction error is controlled within a preset threshold; Second, the dynamic correction mechanism for the heat signal, which divides the full load range of the unit into multiple continuous ranges, calibrates the optimal correction coefficient for each range, and uses linear interpolation to smoothly switch between them, so that the corrected heat signal can accurately adapt to different load conditions; Third, the coordinated control strategy, which dynamically adjusts the control parameters through an adaptive PID algorithm, combined with the optimal ratio control of fuel and feedwater, and the turbine's advance fine-tuning based on prediction deviation, to achieve refined control of the boiler and turbine coordination and ensure control accuracy under complex operating conditions.

[0027] The technical solution of this embodiment has been verified under actual operating conditions and can effectively solve the inherent limitations of traditional DEB control: the main steam pressure fluctuation amplitude is significantly reduced, the deviation is controlled within the preset accuracy range under dynamic operating conditions, and the unit's response speed to AGC load adjustment is greatly improved; the heat correction error breaks through the limitations of traditional methods, realizing refined control and helping the unit reduce coal consumption; at the same time, it is adapted to the complex coupling characteristics of variable load operation of 30%~100% rated load and high parameter units, and can still maintain stable operation under dynamic scenarios such as coal quality fluctuation and efficiency decay, providing reliable technical support for improving the flexibility and energy saving of thermal power units.

[0028] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0029] Example 1 In existing technologies, direct energy balance (DEB) control technology is a classic scheme for coordinated control of boiler and turbine based on the principle of real-time pressure and energy balance. Its core is to achieve rapid response of thermal power unit load and operating parameters by directly linking boiler output and turbine demand, that is, linking heat signal and energy balance signal. It is widely used in coordinated control scenarios of thermal power units.

[0030] Specifically, the core principle of traditional DEB control technology is based on the energy supply and demand balance of boiler turbine. It does not rely on intermediate variables such as indirect conversion of fuel quantity, but directly calculates the energy supply and demand relationship through on-site measured pressure parameters, so that the boiler output tracks the energy demand of the turbine in real time, thereby avoiding the regulation lag problem caused by load and pressure coupling in traditional PID control.

[0031] The key to this technology lies in the calculation of two real-time signals, and all parameters rely on the actual measured pressure collected by the unit's DCS system, without any main steam pressure prediction step: one is the energy balance signal (BD), which serves as a demand-side characterization signal for the turbine, and its calculation formula is as follows: In the formula The main steam pressure, i.e., the pressure before the turbine. Main steam pressure setpoint This refers to the first-stage pressure of the steam turbine, i.e., the regulating stage pressure. , These are the preset adjustment coefficients for the basic operation of the unit, where The ratio directly reflects the opening of the turbine regulating valve, and the differential term is used to capture the rate of change of the turbine's energy demand; the second is the heat signal (HR), which serves as a characterization signal on the boiler output side, and its calculation formula is... In the formula This refers to the steam drum pressure, meaning the separator outlet pressure is used instead of the once-through boiler pressure. This is the real-time differential of the steam drum pressure, used to characterize changes in the boiler's energy storage. Because it maintains a good linear relationship with the main steam flow and the unit load within the range of 40% to 100% of the unit's rated load, it can be used as a substitute parameter for the main steam flow to reflect the actual output of the boiler.

[0032] The control logic architecture of traditional DEB control technology specifically includes: In the main control loop, the boiler main controller uses the measured heat signal HR as the feedback value and the measured energy balance signal BD as the setpoint, and adjusts the boiler's fuel quantity and feedwater flow through a PID controller to ensure that the boiler output tracks the turbine's energy demand; the turbine main controller assists in adjusting the turbine valve opening to quickly respond to unit load commands. At the same time, maintain the main steam pressure Stable, forming a coordinated control with the boiler main control; the signal preprocessing stage only processes the measured pressure parameters ( , , The system performs real-time smoothing and filtering to remove noise, without predicting or anticipating any changes in the main steam pressure.

[0033] Reference Figure 2 The historical trend curves of key pressure parameters during unit operation show the temporal changes of multiple unit parameters obtained in step S1 of this embodiment over a certain period of time: the red curve corresponds to the inlet pressure, i.e., the main steam pressure in step S1, parameter identifier AM055309; the blue curve corresponds to the regulating stage pressure, i.e., the regulating stage pressure in step S1, parameter identifier AM055210; the green curve corresponds to the main steam pressure, i.e., the main steam pressure setpoint in step S1, identifier YLXS; and the purple curve corresponds to the steam drum pressure, i.e., the steam drum pressure in step S1, parameter identifier AM140318. The unit of measurement for each parameter is MPa.

[0034] The above curve data are real-time operating data synchronously acquired by the unit's DCS system. It shows that the main steam pressure fluctuates greatly under the traditional control method and has a continuous deviation from the main steam pressure setpoint. This actual operating characteristic is a direct manifestation of the passive adjustment caused by the lack of pressure prediction and reliance on real-time feedback in the traditional DEB control. It also reflects the necessity of this embodiment to improve parameter stability by predicting the main steam pressure and dynamically correcting the heat signal.

[0035] This embodiment of a unit heat correction method based on main steam pressure prediction includes: Reference Figure 1 S1. Obtain unit parameters, including main steam pressure, main steam pressure setpoint, regulating stage pressure, steam drum pressure, unit load command, fuel flow rate, and feedwater flow rate; The main steam pressure is the turbine inlet pressure, the regulating stage pressure is the turbine first stage pressure, the steam drum pressure is the pressure inside the steam drum of the steam drum boiler, the main steam pressure setpoint is generated according to the unit load command, and the unit load command is issued by the power grid AGC system.

[0036] The main steam pressure is the turbine inlet pressure, collected by a pressure sensor, and directly reflects the turbine inlet steam pressure status. The regulating stage pressure is the turbine's first-stage pressure, collected by a pressure sensor of the same specification. It maintains a good linear relationship with the main steam flow and unit load within the unit's 40%–100% rated load range, and can be used as a substitute parameter for the main steam flow to characterize the steam delivery volume. The drum pressure is the pressure inside the drum of the drum boiler. If the unit is a once-through boiler, the separator outlet pressure is used instead of the drum pressure, which is collected in real-time by a pressure sensor. The main steam pressure setpoint is generated by the unit control system based on the unit load command. The calibration process is based on historical load and pressure matching data to ensure accurate matching between the setpoint and load demand. The unit load command is issued by the power grid AGC system, with the issuance frequency synchronized with the unit parameter acquisition frequency. The fuel flow is collected by an electromagnetic flow monitoring device on the boiler fuel delivery pipeline, and the feedwater flow is collected by a vortex flow monitoring device on the feedwater pipeline. All unit parameters are synchronously acquired through the unit's DCS system at a preset acquisition frequency. The DCS system transmits and caches the acquired parameters in real time to ensure the timeliness of parameter acquisition and the consistency of timestamps for each parameter, providing a reliable data foundation for subsequent data preprocessing and model prediction.

[0037] S2. The unit parameters are preprocessed to obtain standardized data; The process of preprocessing the unit parameters to obtain standardized data includes using an adaptive filtering algorithm to filter the unit parameters and using a normalization method to map the filtered unit parameters to a preset value range.

[0038] The preprocessing process includes outlier removal, filtering and noise reduction, and normalization, with the specific steps as follows: S21. Outlier Removal: For the time series data of each unit parameter, calculate the mean μ and standard deviation σ of the parameter, and identify outliers using the formula |x-μ|>3σ, where x is a single measured value of the parameter. Data identified as outliers are supplemented using linear interpolation to avoid processing bias caused by missing data. S22. Filtering and Denoising: An adaptive filtering algorithm is used to process the unit parameters after outlier removal. The filtering algorithm dynamically adjusts the filtering coefficient according to the parameter fluctuation amplitude. When the parameter fluctuation amplitude is greater than the preset fluctuation threshold, the preferred value is 0.1MPa or the equivalent fluctuation level of the corresponding parameter, and the filtering coefficient is increased to enhance the suppression effect of high-frequency interference signals. When the parameter fluctuation amplitude is less than the preset fluctuation threshold, the filtering coefficient is decreased to retain the dynamic change characteristics of the parameters, thus achieving a balance between interference removal and dynamic feature preservation. S23. Normalization Processing: The Min-Max normalization method is used to map the filtered unit parameters to a preset numerical range of [0,1], eliminating dimensional differences between different parameters. The normalization calculation formula is as follows: ,in, Here, x represents the normalized parameter value, and x represents the measured parameter value after filtering. , These are the maximum and minimum values ​​from the historical operating data of this parameter. Through the above preprocessing steps, the original unit parameters are transformed into standardized data that meets the input requirements of the deep learning prediction model, effectively improving the quality of the model input data.

[0039] S3. Construct a deep learning prediction model, input the standardized data into the deep learning prediction model, and obtain the predicted value of the main steam pressure within the future prediction period; The deep learning prediction model is a Long Short-Term Memory (LSTM) network model. Its hierarchical structure, arranged according to the signal processing flow, consists of an input layer, an embedding layer, three LSTM layers, two fully connected layers, and an output layer. These layers work together to achieve accurate time-series prediction of main steam pressure. The number of hidden units in the three LSTM layers is set to 64, 32, and 16 respectively. This progressively decreasing number of units balances deep feature extraction with redundant information removal, preventing model overfitting. All LSTM layers use the tanh function as their activation function, which effectively introduces nonlinear transformations to adapt to the complex correlation between unit parameters and main steam pressure. The output layer uses a linear activation function to meet the prediction requirements of main steam pressure as a continuous physical quantity, ensuring that the numerical range of the output results is consistent with actual operating conditions.

[0040] Before the model is used for prediction, a training process needs to be completed to ensure prediction accuracy. The specific training steps are as follows: Acquire historical operating data of the target unit under conditions of operation, start-up and shutdown, and coal quality fluctuations within the 30%~100% rated load range. The historical operating data is consistent with the unit parameter type in step S1, and the data duration is not less than 180 days to ensure coverage of all typical operating scenarios of the unit. The preprocessed historical standardized data were divided into training set, validation set and test set in a ratio of 7:2:1. The training set was used for iterative optimization of model parameters, the validation set was used to monitor overfitting, and the test set was used for final accuracy verification. The Adam optimizer was used to train the model, with an initial learning rate of 0.001. A learning rate decay strategy was employed to optimize training efficiency, decreasing the learning rate to 0.9 times the original rate every 50 iterations. Mean squared error (MSE) was used as the loss function, calculated using the following formula: Where N is the number of samples, This is the actual main steam pressure value. To prevent overfitting of the model, an early stopping strategy is adopted during training. Training is stopped when the validation set loss does not decrease for 10 consecutive rounds. After training, the model's prediction accuracy is verified using a test set. The average absolute error between the model's predicted values ​​and the actual values ​​is required to be no more than 0.1 MPa, meeting the accuracy requirements for subsequent heat signal correction. At the same time, an online model update mechanism is set up, collecting the latest unit operation data every 7 days to fine-tune the model parameters, ensuring that the model can maintain stable prediction performance even when the unit's operating status changes, such as efficiency decline or changes in coal quality characteristics.

[0041] Furthermore, the standardized data is input into a deep learning prediction model to obtain the predicted main steam pressure value within the future prediction period, including: S31. Construct an input feature sequence from the standardized data in chronological order. The input feature sequence covers the time-series changes in main steam pressure, main steam pressure setpoint, regulating stage pressure, drum pressure, unit load command, fuel flow, and feedwater flow. The preset time step of the input feature sequence is set to 60 steps, corresponding to a duration of 6 seconds, ensuring that the feature sequence can completely retain the time-series changes in recent unit operating parameters. Each input feature sequence has a dimension of 60×7, where 60 is the time step and 7 represents the types of unit parameters, comprehensively covering the key factors affecting changes in main steam pressure.

[0042] S32. The input feature sequence is mapped to a 32-dimensional preset feature dimension through an embedding layer. This dimension setting has been verified under multiple operating conditions and can enhance the feature expression capability while avoiding dimensional redundancy. Then, the mapped features are input into a 3-layer LSTM layer for deep extraction and time dependency modeling: the first LSTM layer is responsible for capturing the basic time-series change features of each unit parameter, the second LSTM layer extracts the short-term correlation features between parameters, and the third LSTM layer focuses on the complex nonlinear coupling relationship between the future change trends of fuel flow, feedwater flow, regulating stage pressure and main steam pressure, so as to achieve effective decoupling and modeling of the influence of multiple factors.

[0043] S33. The output features of the LSTM layer are compressed through gradient mapping using two fully connected layers in the model: the first fully connected layer maps the 16-dimensional LSTM output features to 8 dimensions, and the second fully connected layer further compresses them to 4 dimensions, effectively reducing feature redundancy and improving computational efficiency; finally, the output layer performs linear mapping transformation to output the predicted value of the main steam pressure within a preset prediction time. The preset prediction time can be configured to 5 to 10 seconds according to the unit's response characteristics, preferably 8 seconds. This allows sufficient adjustment time for subsequent heat signal correction and avoids the decrease in accuracy caused by excessively long prediction times, ensuring that the prediction results can accurately represent the short-term trend of main steam pressure changes.

[0044] S4. Calculate the energy balance signal based on the main steam pressure, the main steam pressure setpoint, and the regulating stage pressure; calculate the original heat signal based on the regulating stage pressure and the steam drum pressure. The energy balance signal is used to represent the turbine energy demand, and the original heat signal is used to represent the boiler heat production capacity. Dynamically correct the original heat signal according to the predicted main steam pressure to obtain the corrected heat signal. The calculation of the energy balance signal based on the main steam pressure, the main steam pressure setpoint, and the regulating stage pressure, and the calculation of the original heat signal based on the regulating stage pressure and the steam drum pressure, include: S41. The formula for calculating the energy balance signal is: ,in, , This is a preset adjustment coefficient for basic unit operation, and its value is calibrated based on unit design parameters, rated load, and historical stable operation data. The value range is 0.05 to 0.2. The value range is 0.1~0.3 to ensure that the sensitivity of the dynamic term to the rate of change of energy demand is adapted to the unit's response characteristics; The main steam pressure is the standardized data after preprocessing in step S2, which is then inversely normalized to its original value. Main steam pressure setpoint For regulating stage pressure, BD is the energy balance signal. In the formula... The steady-state component characterizing the energy demand of the steam turbine reflects the basic demand under the current load; the differential term and It characterizes the dynamic components of demand, and captures the demand fluctuations caused by the rate of change of the regulating valve opening and the rate of change of the set value respectively. The two work together to enable the energy balance signal to fully cover both steady-state and dynamic demand.

[0045] S42. The formula for calculating the original heat signal is: ,in, For the steam drum pressure, The time derivative of the steam drum pressure is used to characterize the change in boiler energy storage. When the boiler heat production exceeds the turbine heat consumption, When the value is positive, energy storage increases; when heat production is less than heat consumption, A negative value indicates reduced energy storage. The regulating pressure is linearly related to the main steam flow rate within the range of 40% to 100% of the rated load, directly representing the boiler's steam output; HR is the raw heat signal, comprehensively reflecting the boiler's real-time heat production capacity and energy storage status.

[0046] , and All calculations employ numerical differentiation, with a preset time interval Δt = 100 ms, consistent with the parameter acquisition frequency in step S1. For any parameter x whose differentiation needs to be calculated, its value at time k is selected. Value at time k−1 Through formula The calculations ensure that the differential results can capture the rate of change of parameters in real time.

[0047] The step of dynamically correcting the original heat signal to obtain the corrected heat signal includes: S43. Based on the predicted main steam pressure value and the set main steam pressure value, calculate the prediction deviation rate between the two. The calculation formula is: ,in, This is the predicted value of steam pressure. The prediction deviation rate is used to quantify the deviation trend of future main steam pressure. When δ>0, it indicates that the future main steam pressure will be higher than the set value, and the boiler heat output needs to be adjusted down in advance. When δ<0, it indicates that the future main steam pressure will be lower than the set value, and the boiler heat output needs to be adjusted up in advance. When δ=0, it indicates that the predicted pressure is consistent with the set value and no correction is needed. The deviation rate can quantify the direction and degree of deviation of future main steam pressure, providing a basis for predictive adjustment.

[0048] S44. Set the preset correction coefficient The preset correction coefficient is used to quantify the degree of influence of the prediction deviation on the original heat signal; its value needs to be adapted to the heat correction sensitivity under different load conditions. The load range calibration method determined the following: The unit's rated load range of 30% to 100% was divided into 6 consecutive preset load ranges, with each range spanning 11.7% of the rated load. A preset coefficient range of 0.3 to 1.2 was set for each preset load range, and candidate coefficient values ​​were selected at 0.1 intervals within this range. Within each preset load range, the unit was stably operated at the midpoint of the range's load. The candidate coefficient values ​​were then substituted into the correction formula sequentially, keeping other control parameters constant. The unit was operated continuously for 30 minutes, and main steam pressure data was collected. The main steam pressure fluctuation amplitude corresponding to each candidate coefficient was calculated. Set a preset fluctuation threshold. Select to make ≤ and The smallest candidate coefficient value is used as the interval. When the unit is running, based on the current actual load, a linear interpolation algorithm is used. Achieve adjacent intervals The smooth switching is achieved, where L is the current load, and La and Lb are the boundary loads of adjacent intervals. , For the corresponding interval This ensures that the correction signal is continuous and without abrupt changes.

[0049] S45. Based on the prediction deviation rate and the preset correction coefficient, the original heat signal is dynamically corrected, and the calculation formula is: ,in, This is a preset correction factor. This is the corrected heat signal.

[0050] The numerical values ​​directly represent the boiler's current heat production capacity and energy storage status; The value includes positive or negative polarity and a specific amplitude. The polarity is determined by the adjustment requirements of the prediction deviation. When the predicted main steam pressure is lower than the set value (δ<0), Take a positive value, so that The calculation result is greater than 1. Compared to The trend is increasing, corresponding to the boiler needing to increase heat production in advance; when the predicted main steam pressure is higher than the set value (δ>0), Take a negative value, so that The calculation result is less than 1. Compared to The heat output is decreasing, and boilers need to reduce their heat production in advance accordingly. The amplitude is determined by load-sharing calibration to ensure that the correction amplitude is adapted to the thermal response characteristics of the unit under different loads, and to avoid insufficient or excessive correction; δ is the prediction deviation rate calculated in step S43, which quantifies the direction and degree of the main steam pressure deviating from the set value in the future, and provides a quantitative basis for the correction direction and amplitude. The corrected heat signal retains the original heat signal's representation of the boiler's current state while incorporating the trend adjustment of main steam pressure prediction, thus playing a role in real-time status and prediction.

[0051] During the correction process, the calculation of HR′ is synchronized with the parameter acquisition frequency in step S1 to ensure the real-time performance of the correction signal; simultaneously, it is combined with the adjacent load intervals in step S44. The linear interpolation switching ensures that the change in HR′ is continuous and without abrupt changes, avoiding boiler combustion fluctuations caused by jumps in the correction coefficient. The corrected HR′ must meet the following requirements: its rate of change does not exceed the maximum allowable rate of change of the unit boiler heat load (preset heat load change threshold), ensuring that the boiler combustion system can respond smoothly to adjustment commands, achieving both predictive and stable control of the main steam pressure and ensuring that the boiler combustion efficiency does not decrease.

[0052] To verify the effectiveness of the correction, the fit between the corrected heat signal HR′ and the energy balance signal BD can be calculated in real time as γ = BD|HR′−BD|. γ is required to be ≤0.05, meaning the fit is no less than 95%. When γ exceeds a preset fit threshold, a trigger is established. The online fine-tuning mechanism further improves the accuracy of corrections, ensuring a dynamic and precise match between boiler output and turbine energy demand.

[0053] The preset correction coefficient needs to be calibrated. The core purpose of calibration is to make the preset correction coefficient adapt to the thermal response characteristics of the unit under different load ranges, ensure that the correction amplitude of the heat signal matches the adjustment requirements of the main steam pressure prediction deviation, and avoid insufficient or excessive correction. The specific calibration steps are as follows: S441. Divide the full load range of the unit from 30% to 100% of its rated load into 6 consecutive preset load ranges. The load span of each preset load range is 11.7% of the rated load, ensuring full coverage of all typical operating load conditions of the unit. Set a corresponding preset coefficient range for each preset load range. The preset coefficient range is -1.2 to 0.3. This range has been verified by the unit's hot commissioning and can cover the adjustment requirements of prediction deviation under different loads. Based on the preset coefficient range, candidate coefficient values ​​are selected at equal intervals of 0.1. The number of candidate coefficient values ​​for each preset load range is 16, ensuring the accuracy of the calibration results.

[0054] S442. Within each preset load range, the unit is stably operated at the intermediate load point of that range, and each of the candidate coefficient values ​​is sequentially used as a preset correction coefficient. Substituting the temporary values ​​into the dynamic correction formula Maintain constant PID control parameters, fuel and feedwater ratios, and other control parameters; set the test duration for each candidate coefficient value to 30 minutes; acquire real-time main steam pressure data according to the preset acquisition frequency in step S1; perform inverse normalization on the acquired main steam pressure data; and then use the formula... Calculate the fluctuation range of the main steam pressure, where This represents the maximum measured value of the main steam pressure during the test period. This represents the minimum measured value of the main steam pressure during the test period.

[0055] S443. Set the preset fluctuation threshold. The preset fluctuation threshold is determined based on the unit's safe and stable operation standards and industry specifications, and represents the maximum allowable fluctuation range of the main steam pressure. The fluctuation range corresponding to all 16 candidate coefficient values ​​within each preset load range is compared. Strictly select those that meet the requirements ≤ The candidate coefficient values ​​are then selected from those that meet the requirements. The smallest value is determined as the preset correction coefficient corresponding to the preset load range. This ensures optimal stability of the main steam pressure within the load range.

[0056] S444. Based on the current actual load L of the unit, a linear interpolation algorithm is used to achieve smooth switching of preset correction coefficients between adjacent preset load intervals. The interpolation calculation formula is as follows: ,in , The boundary load values ​​between two adjacent preset load intervals. , They are respectively , The corresponding preset correction coefficient is calculated through this interpolation, ensuring a continuous transition of the preset correction coefficient to the random group load changes, thus avoiding abrupt changes in the correction coefficient that could affect the corrected heat signal. Sudden changes ensure a smooth response of the boiler combustion system.

[0057] S5. Using the corrected heat signal as a feedback value and the energy balance signal as a set value, the boiler's fuel flow rate and feedwater flow rate are regulated to achieve unit heat correction.

[0058] The purpose of this step is to match the boiler output with the turbine's energy demand through adaptive PID regulation and boiler-turbine coordinated control, while maintaining the optimal combustion ratio of fuel and feedwater. This ensures stable main steam pressure, improves unit operating efficiency, and achieves refined correction of heat distribution.

[0059] The step of using the corrected heat signal as a feedback value and the energy balance signal as a set value to regulate the boiler's fuel flow rate and feedwater flow rate includes: S51. An adaptive PID control algorithm is adopted, with the corrected heat signal as the feedback value and the energy balance signal as the set value. Based on the deviation between the corrected heat signal and the energy balance signal, the proportional coefficient and integral coefficient are dynamically adjusted to adapt the control action to the changes in operating conditions. An adaptive PID control algorithm is adopted, based on the corrected heat signal. The feedback value and the energy balance signal BD are set values, based on The proportional and integral coefficients are dynamically adjusted based on the deviation from the BD value to ensure real-time adaptation of the adjustment action to changes in operating conditions. First, the quantified deviation value is calculated. This value characterizes the degree of fit between heat signals and energy demand; proportionality coefficient. The adjustment formula is Integral coefficient The adjustment formula is ,in =2.5、 =0.1 is the basic adjustment coefficient, based on the unit's rated operating condition calibration; α=0.8 and β=0.3 are preset parameter adjustment coefficients. This is used to limit the adjustment range of the coefficients and avoid over-adjustment. When ε>0.05 and the fit is less than 95%, the algorithm automatically speeds up the coefficient adjustment rate; when ε≤0.05, the coefficients are kept stable to ensure a balance between adjustment accuracy and stability.

[0060] S52. Based on the adjusted adaptive PID control algorithm, the output control signal is transmitted in real time to the boiler fuel delivery actuator and feedwater regulating valve through the unit's DCS system. This regulates the boiler's fuel flow rate F and feedwater flow rate G, achieving a dynamic match between boiler output and turbine energy demand. The output frequency of the control signal is consistent with the parameter acquisition frequency in step S1 to ensure real-time control. Simultaneously, through coordinated control of fuel flow rate and feedwater flow rate, combustion instability caused by single flow rate adjustment is avoided, ensuring that the boiler can maintain efficient combustion under conditions such as variable load and fluctuating coal quality. S53. The turbine main control unit receives unit load commands from the power grid AGC system in real time. and the predicted main steam pressure output in step S3 ,calculate With main steam pressure setpoint absolute value of deviation Set a preset deviation threshold. This threshold is determined based on the unit's safe operation specifications and the main steam pressure control accuracy requirements; when At that time, the turbine main control system fine-tunes the valve opening according to a preset opening gradient, with each adjustment not exceeding 0.5% of the rated opening, and the adjustment interval being 500ms, until... By making predictive fine-tuning adjustments to the turbine valves, the boiler can be assisted in regulating and stabilizing the main steam pressure, further improving the unit's operational stability.

[0061] The regulation of the boiler's fuel flow rate and feedwater flow rate includes: S521. Obtain the optimal ratio data of fuel flow and feedwater flow under different load ranges of the current unit, and establish a load ratio mapping relationship table; Acquire thermal efficiency test data of the target unit within the 30%~100% rated load range. This data covers the optimal ratio of fuel flow to feedwater flow under different load points and main steam pressure deviations. The optimal ratio is defined as the flow ratio that maximizes boiler thermal efficiency and minimizes flue gas heat loss under this operating condition. Following the preset load range division standard in step S44, organize the optimal ratio data according to the structure of load range, main steam pressure deviation range, and optimal fuel / feedwater ratio to establish a load ratio mapping table. For example, within the 30%~41.7% rated load range, when the predicted main steam pressure deviation is -0.3MPa, the corresponding optimal ratio is 1:4.2, i.e., the ratio of fuel flow to feedwater flow.

[0062] S522. Based on the current actual unit load L and the predicted main steam pressure output in step S3. Calculate the preset load range to which the actual load belongs and the corresponding main steam pressure deviation. The optimal load ratio is matched from the load ratio mapping table using a bilinear interpolation algorithm. The interpolation calculation formula is as follows: , in , For the boundary loads of two adjacent intervals, , This corresponds to the main steam pressure deviation boundary value. , , , To ensure that the matching results accurately adapt to the current working conditions, the optimal ratio corresponding to the mapping relationship table is determined.

[0063] S523. During the regulation process, the optimal ratio obtained through matching is used. Based on the corrected heat signal The mixing ratio deviation is dynamically adjusted in real time to account for changes. The allowable range for the mixing ratio deviation is set at ±3%. rate of change At that time, the proportion of water supply flow is increased in a gradient of 0.5%; when At that time, the fuel flow rate percentage is increased in increments of 0.5%; when At the same time, the optimal combustion ratio remains unchanged. Through this dynamic fine-tuning mechanism, while ensuring that the boiler output tracks energy demand, a highly efficient combustion ratio is always maintained, reducing coal consumption and pollutant emissions, and achieving the dual goals of energy saving and stability.

[0064] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages: The technical solution of this application embodiment accurately predicts the future trend of main steam pressure by constructing a long short-term memory network model, and performs predictive dynamic correction of the original heat signal by combining the preset correction coefficient calibrated in the load interval. Then, it dynamically optimizes the control parameters through an adaptive PID adjustment algorithm, coordinates the optimal ratio control of fuel and feedwater and predictive fine-tuning of turbine valves, effectively solving the problems of passive adjustment, large fluctuation of main steam pressure, insufficient heat correction accuracy, and poor adaptability to variable load and high parameter units in traditional DEB control technology. It significantly improves the unit's response speed to AGC load adjustment and the stability of main steam pressure control, greatly reduces heat correction error and coal consumption, and realizes stable, efficient and energy-saving operation of thermal power units within the range of 30% to 100% rated load. It is especially suitable for the coordinated control needs of high parameter units such as supercritical and ultra-supercritical units and coal quality fluctuation scenarios.

[0065] Example 2 Reference Figure 3 A unit heat correction system based on main steam pressure prediction includes: The data acquisition module is configured to acquire unit parameters, including main steam pressure, main steam pressure setpoint, regulating stage pressure, steam drum pressure, unit load command, fuel flow rate, and feedwater flow rate. The data preprocessing module is configured to preprocess the unit parameters to obtain standardized data; The main steam pressure prediction module is configured to construct a deep learning prediction model, input the standardized data into the deep learning prediction model, and obtain the predicted value of the main steam pressure within the future prediction period. The heat signal correction module is configured to calculate an energy balance signal based on the main steam pressure, the main steam pressure setpoint, and the regulating stage pressure; calculate an original heat signal based on the regulating stage pressure and the steam drum pressure; the energy balance signal is used to represent the turbine energy demand; the original heat signal is used to represent the boiler heat production capacity; and dynamically correct the original heat signal according to the predicted main steam pressure to obtain a corrected heat signal. The coordination control module is configured to use the corrected heat signal as a feedback value and the energy balance signal as a set value to regulate the boiler's fuel flow and feedwater flow, thereby achieving unit heat correction.

[0066] A computer-readable storage medium storing a plurality of instructions adapted for loading and execution by a processor of a terminal device of the unit heat correction method based on main steam pressure prediction.

[0067] A terminal device includes a processor and a computer-readable storage medium, the processor being configured to implement various instructions; the computer-readable storage medium being configured to store multiple instructions adapted for loading and execution by the processor of the unit heat correction method based on main steam pressure prediction.

[0068] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0069] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0070] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0071] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0072] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0073] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from the scope of the invention. The spirit and scope of the invention are as follows: Thus, if these modifications and variations of the invention fall within the scope of the claims of the invention and their equivalents, the invention is also intended to include these modifications and variations.

Claims

1. A method for unit heat correction based on main steam pressure prediction, characterized in that, include: The unit parameters are acquired, including main steam pressure, main steam pressure setpoint, regulating stage pressure, steam drum pressure, unit load command, fuel flow rate, and feedwater flow rate. The unit parameters are preprocessed to obtain standardized data; A deep learning prediction model is constructed, and the standardized data is input into the deep learning prediction model to obtain the predicted value of the main steam pressure within the future prediction period. The energy balance signal is calculated based on the main steam pressure, the main steam pressure setpoint, and the regulating stage pressure. The original heat signal is calculated based on the regulating stage pressure and the steam drum pressure. The energy balance signal is used to represent the turbine energy demand, and the original heat signal is used to represent the boiler heat production capacity. The original heat signal is dynamically corrected according to the predicted main steam pressure to obtain the corrected heat signal. Using the corrected heat signal as a feedback value and the energy balance signal as a set value, the boiler's fuel flow rate and feedwater flow rate are regulated to achieve unit heat correction.

2. The unit heat correction method based on main steam pressure prediction according to claim 1, characterized in that, Construct a deep learning prediction model, input the standardized data into the deep learning prediction model, and obtain the predicted value of the main steam pressure within the future prediction period, including: The deep learning prediction model is a long short-term memory network model; The standardized data is arranged in chronological order to construct an input feature sequence; The input feature sequence is mapped to a preset feature dimension through an embedding layer, and the mapped features are deeply extracted and time-dependent modeled through an LSTM layer to obtain the nonlinear coupling relationship between the standardized data in the input feature sequence and the future trend of the main steam pressure. The output features of the LSTM layer are compressed by gradient mapping through a fully connected layer, and then linearly mapped and transformed by the output layer to output the predicted value of the main steam pressure within the preset prediction time.

3. The unit heat correction method based on main steam pressure prediction according to claim 1, characterized in that, The calculation of the energy balance signal based on the main steam pressure, the main steam pressure setpoint, and the regulating stage pressure, and the calculation of the original heat signal based on the regulating stage pressure and the steam drum pressure, include: The formula for calculating the energy balance signal is: ,in, , The preset adjustment coefficients for basic unit operation. Main steam pressure, Main steam pressure setpoint For regulating stage pressure, This is an energy balance signal. The formula for calculating the original heat signal is: ,in, Used to indicate changes in the energy storage of a boiler. For the steam drum pressure, This is the original heat signal.

4. The unit heat correction method based on main steam pressure prediction according to claim 3, characterized in that, The step of dynamically correcting the original heat signal to obtain the corrected heat signal includes: Based on the predicted main steam pressure value and the set main steam pressure value, the prediction deviation rate between the two is calculated using the following formula: ,in, This is the predicted value of steam pressure. The predicted deviation rate is used to quantify the deviation trend of future main steam pressure. A preset correction coefficient is set, which is used to quantify the degree of influence of the prediction deviation on the original heat signal; The original heat signal is dynamically corrected based on the prediction deviation rate and the preset correction coefficient, and the calculation formula is as follows: ,in, This is a preset correction factor. This is the corrected heat signal.

5. The unit heat correction method based on main steam pressure prediction according to claim 4, characterized in that, The preset correction coefficient needs to be calibrated, and the calibration of the preset correction coefficient includes: The unit's full load range is divided into multiple consecutive preset load ranges. A corresponding preset coefficient range is set for each preset load range, and multiple candidate coefficient values ​​are selected at equal intervals based on the preset coefficient range. Within each preset load range, the unit is stably operated at the intermediate load point of that range, and each of the candidate coefficient values ​​is sequentially used as a preset correction coefficient. Substituting the temporary values ​​into the dynamic correction formula Keep other control parameters constant, and calculate the fluctuation range of the main steam pressure based on the acquired main steam pressure data; A preset fluctuation threshold is set, which is the maximum main steam pressure fluctuation range allowed for safe and stable operation of the unit. The fluctuation ranges of all candidate coefficient values ​​in each preset load range are compared, and the candidate coefficient value that satisfies the condition that the fluctuation range is less than the preset fluctuation threshold and has the smallest fluctuation range is selected as the preset correction coefficient corresponding to the preset load range. Based on the current actual load of the unit, a linear interpolation algorithm is used to achieve smooth switching of preset correction coefficients between adjacent preset load intervals.

6. The unit heat correction method based on main steam pressure prediction according to claim 1, characterized in that, The step of using the corrected heat signal as a feedback value and the energy balance signal as a set value to regulate the boiler's fuel flow rate and feedwater flow rate includes: An adaptive PID control algorithm is adopted, with the corrected heat signal as the feedback value and the energy balance signal as the set value. The proportional coefficient and integral coefficient are dynamically adjusted based on the deviation between the corrected heat signal and the energy balance signal, so that the control action adapts to the changes in operating conditions. Based on the adjusted adaptive PID control algorithm, the output control signal is used to regulate the boiler's fuel flow and feedwater flow, so that the boiler output and the turbine's energy demand are dynamically matched. The turbine main control unit receives the unit load command and the predicted value of the main steam pressure. When the deviation between the predicted value of the main steam pressure and the set value of the main steam pressure exceeds the preset deviation threshold, the turbine valve opening is adjusted to help stabilize the main steam pressure.

7. The unit heat correction method based on main steam pressure prediction according to claim 6, characterized in that, The regulation of the boiler's fuel flow rate and feedwater flow rate includes: Obtain the optimal ratio data of fuel flow and feedwater flow under different load ranges of the current unit, and establish a load ratio mapping table; Based on the current actual load of the unit and the predicted main steam pressure, the corresponding optimal ratio is matched from the mapping table using an interpolation algorithm; During the control process, the optimal ratio is used as a benchmark, and the ratio deviation is dynamically fine-tuned according to the real-time changes in the corrected heat signal.

8. A unit heat correction system based on main steam pressure prediction, characterized in that, The unit heat correction method based on main steam pressure prediction according to any one of claims 1-7 includes: The data acquisition module is configured to acquire unit parameters, including main steam pressure, main steam pressure setpoint, regulating stage pressure, steam drum pressure, unit load command, fuel flow rate, and feedwater flow rate. The data preprocessing module is configured to preprocess the unit parameters to obtain standardized data; The main steam pressure prediction module is configured to construct a deep learning prediction model, input the standardized data into the deep learning prediction model, and obtain the predicted value of the main steam pressure within the future prediction period. The heat signal correction module is configured to calculate an energy balance signal based on the main steam pressure, the main steam pressure setpoint, and the regulating stage pressure; calculate an original heat signal based on the regulating stage pressure and the steam drum pressure; the energy balance signal is used to represent the turbine energy demand; the original heat signal is used to represent the boiler heat production capacity; and dynamically correct the original heat signal according to the predicted main steam pressure to obtain a corrected heat signal. The coordination control module is configured to use the corrected heat signal as a feedback value and the energy balance signal as a set value to regulate the boiler's fuel flow and feedwater flow, thereby achieving unit heat correction.

9. A computer-readable storage medium storing a plurality of instructions, characterized in that, The instructions are adapted to be loaded and executed by the processor of the terminal device as described in claim 1, which is a unit heat correction method based on main steam pressure prediction.

10. A terminal device, comprising a processor and a computer-readable storage medium, wherein the processor is configured to implement instructions; and the computer-readable storage medium is configured to store multiple instructions, characterized in that, The instructions are adapted to be loaded by a processor and executed as described in claim 1, which is a unit heat correction method based on main steam pressure prediction.

Images