A minimum flow control system for a steam turbine feed pump

Abstract

The present application relates to the field of steam turbine, disclose a kind of steam turbine feedwater pump minimum flow control system, for by combining feedwater pump minimum flow control with deaerator pressure control organically, realize the collaborative optimization of two systems. Including comparison prediction sequence and feedwater flow set value and deaerator pressure set value obtain collaborative error vector, generate sliding mode control amount, combined with system operation constraint condition solution obtain regulating valve optimal opening degree instruction. The present application deeply couples feedwater thermal deaeration process and feedwater pump minimum flow control, accurately describes system dynamic characteristics, realizes the collaborative optimization of feedwater flow and deaerator pressure by multi-step prediction and adaptive sliding mode control, significantly improves the control precision of feedwater system.

Description

Technical Field

[0001] This invention relates to the field of steam turbines, and more particularly to a minimum flow control system for a steam turbine feedwater pump. Background Technology

[0002] In thermal power generation systems, minimum flow control of the turbine feedwater pump is a crucial aspect of ensuring the safe operation of the boiler. Minimum flow control aims to prevent excessively high fluid temperatures within the pump, which could lead to cavitation, while simultaneously maintaining stable system operation. Traditional control methods primarily rely on PID control strategies, adjusting the recirculation valve opening to maintain the minimum flow rate. However, this approach has significant limitations, particularly in its coordination with the deaerator system.

[0003] In existing technologies, the thermal deoxygenation process of a water supply system and the minimum flow control of the water supply pump are often treated as two independent systems. This separate control strategy leads to several significant problems:

[0004] Traditional control methods neglect the dynamic coupling characteristics between the feedwater system and the deaerator system. In actual operation, changes in feedwater flow directly affect the pressure balance within the deaerator, while fluctuations in deaerator pressure, in turn, affect the operating conditions of the feedwater pump. This two-way coupling relationship has not been adequately considered or effectively addressed in existing control schemes.

[0005] Existing minimum flow control systems lack effective handling of system operational constraints. In practical engineering, feedwater pump regulating valves have physical constraints such as mechanical limits and operating rates, while the system also needs to meet various safe operating boundary conditions. Traditional control methods often treat these constraints as independent protection functions rather than integrating them into the design of the control algorithm, which may lead to a decrease in system performance under boundary conditions.

[0006] Existing technologies for enhancing safety mostly employ passive protection strategies, meaning protective actions are only taken after an anomaly is detected. This reactive approach may not be effective in preventing accidents, especially under rapidly changing operating conditions.

[0007] Therefore, we propose a minimum flow control system for turbine feedwater pumps to solve the above problems. Summary of the Invention

[0008] This invention provides a minimum flow control system for a steam turbine feedwater pump, which combines the minimum flow control of the feedwater pump with the pressure control of the deaerator to achieve synergistic optimization of the two systems.

[0009] The first aspect of this invention provides a minimum flow control system for a steam turbine feedwater pump, comprising: a data acquisition module for acquiring real-time operating parameters of the feedwater system and generating a set of real-time operating parameters; a prediction module for inputting the set of real-time operating parameters into a preset fractional-order state-space model to generate a multi-step prediction sequence; a processing module for comparing the multi-step prediction sequence with preset feedwater flow rate setpoints and deaerator pressure setpoints to generate a cooperative error vector; a setting module for processing the cooperative error vector by inputting it into an adaptive fractional-order sliding surface to generate a sliding mode control quantity; and an allocation module for solving for the optimal opening command of the regulating valve based on the sliding mode control quantity and preset system operating constraints.

[0010] Optionally, in a first implementation of the first aspect of the present invention, the method includes: fusing the real-time operating parameter set with the historical state data sequence to generate an extended state vector; inputting the extended state vector into a fractional-order state transition matrix to calculate a state evolution sequence for multiple future sampling times; performing output extraction processing based on the state evolution sequence to separate the feedwater flow evolution component and the deaerator pressure evolution component, generating an uncorrected prediction sequence; performing dynamic error correction on the uncorrected prediction sequence to generate a corrected prediction sequence; and truncating and recombining the corrected prediction sequence according to the prediction time domain length to generate a multi-step prediction sequence.

[0011] Optionally, in a second implementation of the first aspect of the present invention, the step of inputting the extended state vector into a fractional-order state transition matrix to calculate the state evolution sequence at multiple future sampling times includes: extracting typical operating modes of the water supply system under different load conditions from a historical operating database to generate a historical operating mode library; performing similarity matching between the extended state vector and the typical operating modes in the historical operating mode library to identify the feature parameters of the current operating mode and generate a mode matching result; based on the mode matching result, selecting a corresponding fractional-order differential operator from a preset fractional-order operator library to generate fractional-order calculation parameters adapted to the current operating state; using the fractional-order calculation parameters to perform fractional-order state transition calculation on the extended state vector to obtain the evolution trajectory of the system state at multiple future sampling times and generate an initial state evolution sequence; and performing perturbation compensation on the initial state evolution sequence in conjunction with real-time acquired perturbation feature parameters to generate a refined state evolution sequence.

[0012] Optionally, in a third implementation of the first aspect of the present invention, the method includes: extracting a feedwater flow rate reference setpoint and a deaerator pressure reference setpoint from a preset load-pressure characteristic curve based on the current unit load parameters and deaerator operating conditions, and generating a setpoint sequence for multiple future sampling times; comparing the predicted feedwater flow rate in the multi-step prediction sequence with the feedwater flow rate reference setpoint in the setpoint sequence point by point, and simultaneously comparing the predicted deaerator pressure with the deaerator pressure reference setpoint point by point, to generate a preliminary error sequence; performing dynamic coupling analysis based on the preliminary error sequence to identify the coupling relationship between the feedwater flow rate error and the deaerator pressure error, and generating an error coupling feature matrix; performing weighted reconstruction of the preliminary error sequence based on the error coupling feature matrix to generate a dynamic error sequence; and processing the dynamic error sequence to generate a cooperative error vector.

[0013] Optionally, in the fourth implementation of the first aspect of the present invention, the method includes: obtaining fractional-order parameters, convergence rate parameters, and coupling weight parameters from a preset sliding mode parameter configuration table according to the current operating conditions of the system, and generating an initial sliding mode surface parameter configuration; calculating the fractional-order differential sequences of the feedwater flow tracking error and the deaerator pressure tracking error according to the cooperative error vector, and generating an error differential feature set; calculating the basic sliding mode surface quantity through the sliding mode surface dynamic equation based on the error differential feature set and the initial sliding mode surface parameter configuration; performing online adaptive adjustment of the initial sliding mode surface parameter configuration according to the system state change rate, and generating real-time optimized sliding mode surface parameters; and reconstructing the basic sliding mode surface quantity using the real-time optimized sliding mode surface parameters to generate a sliding mode control quantity.

[0014] Optionally, in a fifth implementation of the first aspect of the invention, the basic sliding surface quantity s0 is calculated based on the error differential feature set and initial parameters:

[0015]

[0016] in, This is to account for the error in water supply flow rate; This refers to the deaerator pressure error. λ is a fractional differential operator; λ is the convergence rate parameter; w is the coupling weight.

[0017] Optionally, in a sixth implementation of the first aspect of the present invention, the step of online adaptively adjusting the initial parameter configuration of the sliding surface according to the system state change rate to generate real-time optimized sliding surface parameters includes: real-time monitoring of the state change characteristics of the water supply system to generate a system state change rate feature set; analyzing the matching relationship between the system dynamic response characteristics and the sliding surface parameters based on the system state change rate feature set to generate a parameter adjustment strategy; coordinating the adjustment of fractional-order parameters, convergence rate parameters, and coupling weight parameters according to the parameter adjustment strategy to generate initially adjusted sliding surface parameters; evaluating the stability margin of the initially adjusted sliding surface parameters through a stability verification mechanism to generate parameter stability verification results; and performing final optimization of the initially adjusted sliding surface parameters based on the parameter stability verification results to generate real-time optimized sliding surface parameters.

[0018] Optionally, in the seventh implementation of the first aspect of the present invention, the method includes: extracting the mechanical limit constraints, action rate constraints, and safe operation constraints of the feedwater pump regulating valve to generate a multi-dimensional system constraint set; iteratively solving the problem using a multi-objective optimization algorithm based on the sliding mode control quantity and the multi-dimensional system constraint set to generate a preliminary optimized opening sequence; dynamically smoothing the preliminary optimized opening sequence to eliminate abrupt changes in control commands and generating a smooth opening command sequence; performing feedforward compensation based on the nonlinear flow characteristics of the regulating valve according to the smooth opening command sequence to generate a compensated opening command sequence; extracting the optimal opening command at the current moment from the compensated opening command sequence, converting it into an analog control signal through a digital-to-analog converter, and sending the signal to the feedwater pump regulating valve actuator to achieve coordinated control of the minimum flow rate of the feedwater pump and the deaerator pressure.

[0019] Optionally, in the eighth implementation of the first aspect of the present invention, a protection module is further included: a system real-time status monitoring set is generated by real-time monitoring of key operating parameters of the water pump system; the system real-time status monitoring set is compared with a preset safe operating threshold to identify the degree of abnormal deviation of the system operating state and generate a system safety status assessment result; based on the system safety status assessment result, the protective constraint boundary of the control system is dynamically adjusted to generate an adaptive safety constraint boundary; when the system operating state is detected to be close to the safety boundary, a preventive control mechanism is activated to generate a protective control compensation quantity; the protective control compensation quantity is integrated into the main control loop to enhance the system safety margin without interrupting the normal control process, thereby achieving enhanced safe operation of the water pump minimum flow control system.

[0020] The mechanism of this invention is as follows: by constructing a deeply coupled system of feedwater thermal deoxygenation process and feedwater pump flow control, a fractional-order state-space model is used to accurately describe the dynamic characteristics of the system, and combined with an adaptive sliding mode predictive control strategy, the coordinated optimization control of the thermal and hydraulic parameters of the feedwater system is realized.

[0021] Beneficial effects: By deeply coupling the thermal deoxygenation process of feedwater with the minimum flow control of the feedwater pump, a collaborative control architecture based on a fractional-order state-space model was established. Through dynamic coupling analysis, the intrinsic relationship between the two was identified, achieving true multivariate collaborative optimization. This effectively solved the long-standing problem of mutual interference between systems and significantly improved the overall control quality.

[0022] By monitoring the system's state change characteristics in real time and dynamically adjusting the sliding surface parameters, the system can maintain excellent performance under different operating conditions, enabling the control system to effectively cope with the inherent nonlinear and time-varying characteristics of the water supply system and overcome the shortcomings of traditional fixed parameter controllers with poor adaptability.

[0023] By incorporating physical constraints such as the mechanical limit of the control valve, the action rate, and the system's safe operating boundary into the control algorithm design, and solving for the optimal control command that satisfies all constraints through a multi-objective optimization algorithm, the system's safety is guaranteed and its control performance is improved. This proactive constraint management strategy has significant advantages over the traditional passive protection method.

[0024] By conducting real-time status monitoring and safety boundary assessment, protective control compensation is initiated before the system approaches safety limits. This proactive safety strategy can effectively prevent accidents and significantly improve the operational reliability of the system. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of one embodiment of the turbine feedwater pump minimum flow control system in this invention.

[0026] Figure 2 This is a schematic diagram of another embodiment of the turbine feedwater pump minimum flow control system in this invention;

[0027] Figure 3 This is a schematic diagram of the later stage of the minimum flow control system for the steam turbine feedwater pump in an embodiment of the present invention, from sliding mode control to safety enhancement control.

[0028] Figure 4 This is a schematic diagram of one embodiment of the turbine feedwater pump minimum flow control device in the present invention. Detailed Implementation

[0029] This invention provides a turbine feedwater pump minimum flow control system, which organically combines feedwater pump minimum flow control with deaerator pressure control to achieve synergistic optimization of the two systems. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0030] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 One embodiment of the turbine feedwater pump minimum flow control system of the present invention includes:

[0031] 101. Acquisition module, used to acquire real-time operating parameters of the water supply system through sensors, and generate a set of real-time operating parameters including water supply flow rate, pump outlet pressure, pump speed and deaerator pressure;

[0032] It is understood that the executing entity of this invention can be a minimum flow control device for a steam turbine feedwater pump, or it can be a terminal or a server; no specific limitation is made here. This embodiment of the invention will be described using a server as an example.

[0033] It should be noted that, taking a 1000MW ultra-supercritical unit's steam-driven feedwater pump system as an example, the unit configuration includes two steam-driven feedwater pumps (each handling 50% of the load) and one electric standby pump. Sensor data is integrated and processed through a distributed control system (DCS).

[0034] The sensor layout for the water supply system is as follows: Water supply flow measurement: Standard flow nozzles (ISO 5167 standard) are installed on the main water supply pipeline, and orifice flow meters are installed at the inlet of each water supply pump. The flow signal is temperature-corrected (water supply temperature corrected to standard conditions) to reduce errors caused by changes in medium temperature. Real-time data is transmitted to the DCS as a 4-20mA analog signal. Pump outlet pressure measurement: A pressure transmitter (range 0-25MPa) with an accuracy of ±0.1% is installed before the water supply pump outlet valve. The pressure signal is linked to the pump's safe operating range (upper limit characteristic curve) to prevent overpressure. Pump speed measurement: A magnetoelectric speed sensor (detection range 0-6000rpm) is installed at the shaft end of the small turbine. The signal is verified by a WOODWARD 505 controller or MEH system to avoid data distortion caused by slip. Deaerator pressure measurement: An absolute pressure transmitter (range 0-1.2MPa) is installed on the top of the deaerator. It is compensated in conjunction with the deaerator water level signal (differential pressure measured through the balance vessel) to ensure that the pressure measurement reflects the true thermal state.

[0035] The unit operates at 75% load. The real-time parameters collected by the sensors are as follows (simulated data): Feedwater flow rate: The main feedwater flow rate measured by the nozzle is 650 t / h (corresponding to boiler evaporation rate). The inlet flow rate of a single steam pump is 325 t / h (calculated using orifice plate differential pressure, with an error of <±0.5% after temperature correction). Pump outlet pressure: The pump outlet transmitter displays 15.8 MPa, compared with the upper limit of the safe operating range (16.2 MPa) at the current pump speed, ensuring that it is not exceeded. Pump speed: The small steam turbine speed is 5100 rpm, and the deviation between the speed sensor feedback signal and the DCS setpoint is <10 rpm, indicating stable speed control. Deaerator pressure: The deaerator pressure is 0.75 MPa (corresponding to a saturation temperature of 170℃), synchronously verified with the deaerator water level signal (2.5 m) to eliminate false fluctuations.

[0036] The DCS scans sensor signals every 500ms and performs the following steps to generate a parameter set: Signal verification: Triple redundancy verification is performed on each parameter (three independent transmitters are used for water flow), outlier values ​​(data with a change rate > 5% / s) are removed, and the median value is taken as the valid value. Unit standardization: Flow rate is standardized to t / h, pressure to MPa, and rotational speed to rpm, and a timestamp (2025-11-10 10:30:00.000) is added. Data packaging: The data is temporarily stored in a structured format (array in memory) containing the fields: {flow:650,outlet_pressure:15.8,rpm:5100,deaerator_pressure:0.75}, for subsequent use by the prediction model.

[0037] 102. Prediction module, used to input the set of real-time operating parameters into a preset fractional-order state-space model to generate a multi-step prediction sequence containing predicted values ​​of feedwater flow rate and deaerator pressure.

[0038] It should be noted that in the minimum flow control of the steam-driven feedwater pump of a 1000MW ultra-supercritical unit, the dynamic characteristics of the feedwater system are described by fractional calculus, which can more accurately capture the memory and historical dependence of parameter changes.

[0039] The fractional-order state-space model was pre-trained based on historical unit operating data. The model's state variables include fractional derivatives of feedwater flow rate, deaerator pressure, pump outlet pressure, and pump speed. The system matrix was identified using data from typical operating conditions such as unit start-up and shutdown, and load variations. The model parameters were set as follows: fractional order α = 0.85 (determined through optimization using historical data), prediction step size of 5 steps (each step with a 2-second time interval), and 4-dimensional state equations (corresponding to 4 input parameters). During model initialization, the pre-trained weight matrix was loaded, and the initial state vector was set to zero.

[0040] The system receives the real-time parameter set acquired in step 101 (feedwater flow rate 650 t / h, pump outlet pressure 15.8 MPa, pump speed 5100 rpm, deaerator pressure 0.75 MPa). First, the parameters are standardized to eliminate the influence of dimensions: the feedwater flow rate is converted to a per-unit value (baseline value 1000 t / h), the pump outlet pressure is converted to a baseline value of 20 MPa, the pump speed to a baseline value of 6000 rpm, and the deaerator pressure to a baseline value of 1.0 MPa. After standardization, the parameter set becomes a vector form: 0.65, 0.79, 0.85, 0.75. This vector serves as the model input and is concatenated with the historical state vector (storing data from the previous 5 seconds) to form the extended state input.

[0041] The multi-step prediction sequence generation process involves fractional-order state updates: the model is based on the fractional-order accumulation principle, weighting the input states. New data has a higher weight (current data has a weight coefficient of 0.6, while data from the previous time step has a weight of 0.3) to reflect the recent dependence of the system's dynamics. The future trend of the state variables is calculated using fractional-order differential operators, showing that the rate of change in water flow is influenced by the cumulative effect of historical flow values.

[0042] State-space recursion: Discretized state equations are used for step-by-step prediction. The current state vector is used to recursively predict the future state through the system matrix (describing the coupling relationship between parameters, such as the transfer gain of pump speed to feed flow) and the input matrix (reflecting the impact of disturbances). Each step of the recursion considers system noise (random fluctuations caused by pump vibration) and corrects prediction bias using Kalman filtering.

[0043] Output sequence generation: After 5 recursive steps, the predicted sequence for the next 10 seconds is output: Predicted feedwater flow rate: Steps 1-5 are 651 t / h, 652 t / h, 653 t / h, 654 t / h, and 655 t / h respectively (due to the slow increase in load, the flow rate shows a linear growth trend). Predicted deaerator pressure: Steps 1-5 are 0.751 MPa, 0.752 MPa, 0.752 MPa, 0.753 MPa, and 0.753 MPa respectively (due to the gradual change in pressure due to thermal inertia). The predicted sequence is timestamped to form a structured multi-step output for use by the subsequent comparison module.

[0044] The model incorporates a residual detection mechanism: it compares the predicted value with a sliding window of real-time data (the 10 most recent sampling points). If the mean absolute error exceeds a threshold (water flow error > 2 t / h), it triggers fine-tuning of the model parameters (adjusting the fractional order to 0.88). Simultaneously, it performs physical constraint verification on the prediction results (non-negative flow rate, pressure not exceeding the safety limit) and eliminates unreasonable values.

[0045] 103. Processing module, used to compare the multi-step prediction sequence with the preset feedwater flow rate setpoint and deaerator pressure setpoint, and generate a collaborative error vector including feedwater flow rate tracking error and deaerator pressure tracking error;

[0046] It should be noted that in the minimum flow control of the steam-driven feedwater pump of the 1000MW ultra-supercritical unit, the core of step 103 is to compare the multi-step prediction sequence generated by the fractional state-space model with the preset set value to generate a cooperative error vector.

[0047] The feedwater flow rate setpoint and deaerator pressure setpoint are dynamically preset based on the unit's real-time load conditions. Under 75% load (corresponding to a boiler evaporation rate of approximately 650 t / h): Feedwater flow rate setpoint: Based on the boiler main control command (generated from the load-flow function relationship, ensuring it is not lower than 25% ECR of the boiler's minimum flow rate requirement), it is set to 650 t / h. Deaerator pressure setpoint: Based on the deaerator sliding pressure operation curve (to maintain deaeration effect and prevent cavitation), it is set to 0.75 MPa. The setpoints are refreshed every 2 seconds and constrained within safe ranges by the limiting module in the DCS (feedwater flow rate setpoint not lower than 280 t / h, deaerator pressure setpoint not higher than 0.9 MPa).

[0048] The multi-step prediction sequence output from step 102 includes predicted values ​​for five steps (each step 2 seconds apart) within the next 10 seconds: Feedwater flow prediction sequence: 651 t / h, 652 t / h, 653 t / h, 654 t / h, 655 t / h, reflecting a slow upward trend in load. Deaerator pressure prediction sequence: 0.751 MPa, 0.752 MPa, 0.752 MPa, 0.753 MPa, 0.753 MPa, showing a gradual change due to thermal inertia.

[0049] Compare the predicted values ​​with the set values ​​point by point and calculate the absolute error: Water flow tracking error calculation: Step 1 error: 651t / h - 650t / h = 1t / h; Step 2 error: 652t / h - 650t / h = 2t / h; Step 3 error: 653t / h - 650t / h = 3t / h; Step 4 error: 654t / h - 650t / h = 4t / h; Step 5 error: 655t / h - 650t / h = 5t / h; Generate flow error sequence: 1, 2, 3, 4, 5 (unit: t / h).

[0050] Deaerator pressure tracking error calculation: Step 1 error: 0.751MPa - 0.75MPa = 0.001MPa; Step 2 error: 0.752MPa - 0.75MPa = 0.002MPa; Step 3 error: 0.752MPa - 0.75MPa = 0.002MPa; Step 4 error: 0.753MPa - 0.75MPa = 0.003MPa; Step 5 error: 0.753MPa - 0.75MPa = 0.003MPa; Generating pressure error sequence: 0.001, 0.002, 0.002, 0.003, 0.003 (unit: MPa). Error sign convention: Positive error indicates that the predicted value is higher than the set value, requiring a reduction in feedwater flow or pressure stabilization; negative error indicates the opposite.

[0051] The above error sequences are combined into a cooperative error vector E, the structure of which is: E=[feed water flow error sequence, deaerator pressure error sequence]=[1, 2, 3, 4, 5, 0.001, 0.002, 0.002, 0.003, 0.003];

[0052] The vector has 10 dimensions (5-dimensional flow error + 5-dimensional pressure error), is stored in a memory array, and is timestamped. The vector content is transmitted to the adaptive fractional sliding surface module in real time, and the increasing trend of the error (flow error increases from 1t / h to 5t / h) indicates that the system needs to strengthen overshoot suppression.

[0053] 104. Setting module, used to input the cooperative error vector into the adaptive fractional sliding surface for processing, to generate a sliding control quantity that simultaneously includes the feedwater flow control component and the deaerator pressure stability component;

[0054] It should be noted that in the minimum flow control of the steam-driven feedwater pump of the 1000MW ultra-supercritical unit, the core of step 104 is to input the cooperative error vector into the adaptive fractional sliding surface for processing to generate the sliding control quantity.

[0055] The sliding surface is designed as a bivariate collaborative structure, corresponding to the feedwater flow error and the deaerator pressure error, respectively. Its core parameters include: Fractional order: set to 0.9 (determined through historical data optimization), used to weight the historical cumulative effect of the error (the error change trend over the past 10 seconds), enhancing the ability to capture the system's nonlinear dynamics. Adaptive boundary layer thickness: the initial thickness is set to 0.05, designed to dynamically adjust with the absolute value of the error (automatically increasing to 0.1 when the error is large to suppress chattering, and converging to 0.01 when the error is small to improve accuracy). Sliding surface parameters: the weighting coefficient for the flow error sliding surface is 0.7, and the weighting coefficient for the pressure error sliding surface is 0.3, reflecting the dominance of flow control.

[0056] The received collaborative error vector E=[1, 2, 3, 4, 5, 0.001, 0.002, 0.002, 0.003, 0.003] (units: t / h and MPa) generated in step 103 is normalized: the feedwater flow rate error sequence is divided by the reference value of 10t / h to obtain the normalized value [0.1, 0.2, 0.3, 0.4, 0.5]. The deaerator pressure error sequence is divided by the reference value of 0.01MPa to obtain the normalized value [0.1, 0.2, 0.2, 0.3, 0.3]. After normalization, the vector is split into flow rate error sub-vectors and pressure error sub-vectors, which are input into two independent channels of the sliding surface for processing.

[0057] Fractional error accumulation calculation: Based on fractional calculus, the error sequence is weighted and accumulated. Recent errors have higher weights (0.6 for the current time step and 0.3 for the previous time step), while historical errors decay exponentially. The fractional accumulation result for flow rate error is: the weighted sum of steps 1 to 5 = 0.1 × 0.6 + 0.2 × 0.5 + 0.3 × 0.4 + 0.4 × 0.3 + 0.5 × 0.2 = 0.45 (dimensionless). The fractional accumulation result for pressure error is: (Dimensionless).

[0058] Adaptive sliding surface update: The sliding surface dynamically adjusts the boundary layer thickness based on the error. When the cumulative flow error of 0.45 exceeds the threshold of 0.3, the boundary layer thickness adaptively increases from 0.05 to 0.08 to smoothly control the output; the cumulative pressure error of 0.35 triggers the thickness adjustment to 0.06. The sliding surface output consists of two intermediate values: flow channel output -0.52 (negative values ​​indicate that the flow rate needs to be reduced), and pressure channel output +0.28 (positive values ​​indicate that the pressure needs to be increased for stabilization).

[0059] Control quantity synthesis and component allocation: Intermediate quantities are synthesized into a total sliding mode control quantity according to their weights. (Dimensionless). Decomposed into two components: water supply flow control component: (Negative values ​​correspond to reducing the opening of the feedwater regulating valve to suppress excessive flow rate increase). Deaerator pressure stabilization component: (Negative values ​​correspond to fine-tuning the deaerator pressure setpoint to prevent overpressure). The final sliding mode control value is stored in vector form: [-0.21, -0.07].

[0060] The system monitors the chattering intensity (rate of change of control quantity) of the sliding surface output in real time. If the rate of change exceeds the threshold (0.1 / s), the boundary layer thickness is automatically increased to 0.1, and parameter recalibration is triggered (the fractional order is temporarily adjusted to 0.85) to ensure smooth and reliable control. At the same time, physical constraint verification is performed on the control components (flow component not exceeding ±0.5, pressure component not exceeding ±0.2) to avoid over-adjustment of the actuator.

[0061] 105. The allocation module is used to solve for the optimal opening command of the regulating valve based on the sliding mode control quantity and the preset system operation constraints, and send the command to the feedwater pump regulating valve actuator to realize the coordinated control of the minimum flow rate of the feedwater pump and the deaerator pressure.

[0062] It should be noted that in the minimum flow control of the steam-driven feedwater pump of the 1000MW ultra-supercritical unit, the core of step 105 is based on the sliding mode control quantity (feedwater flow control component -0.21 and deaerator pressure stability component -0.07 output from the previous step).

[0063] The constraints are pre-set according to the unit's safe operation standards: Valve characteristic constraints: The feedwater regulating valve adopts an equal percentage flow characteristic (the slope of the characteristic curve is proportional to the opening degree), ensuring smooth regulation at small opening degrees and sensitive response at large opening degrees. The valve opening range is limited to 10%~90% (fully closed to fully open) to avoid oscillation at small opening degrees or uncontrolled operation at large opening degrees. Flow safety constraints: The minimum flow rate of a single steam-driven feedwater pump is set at 320t / h (determined based on the pump characteristic curve) to prevent low-flow cavitation; the maximum flow rate limit is 900t / h (corresponding to the upper limit characteristic curve of the pump). Pressure constraints: The allowable pressure fluctuation range of the deaerator is 0.7~0.9MPa (sliding pressure operation requirement), and the pump outlet pressure limit is 10~20MPa (to avoid overpressure or insufficient pressure). Cavitation prevention constraints: The outlet pressure of the regulating valve must be higher than the saturation pressure corresponding to the current water temperature (saturation pressure is 1.1MPa at 184℃), with a safety margin of 0.3MPa.

[0064] After receiving the sliding mode control quantity ([-0.21, -0.07]), solve it according to the following steps: Multi-objective weight allocation: the feedwater flow control weight is 0.7 (dominant objective, ensuring the flow rate is not lower than the minimum limit), and the deaerator pressure stability weight is 0.3 (auxiliary objective). Calculate the comprehensive control quantity: (Dimensionless). Negative values ​​indicate that the flow rate needs to be reduced to suppress overshoot.

[0065] Constraint optimization solution: Using the current real-time parameters as the initial state (feed flow rate 650 t / h, deaerator pressure 0.75 MPa, pump outlet pressure 15.8 MPa). Searching for the optimal opening within the constraints: First, convert the comprehensive control variables into flow adjustment variables, i.e. t / h (rounded to one decimal place). This value is higher than the minimum flow rate of 320 t / h, which complies with safety constraints.

[0066] Based on the valve flow characteristic curve (equal percentage type), the opening command is calculated: the target flow rate of 641.6 t / h corresponds to a valve opening of approximately 58% (determined by looking up a table; at a fully open flow rate of 900 t / h, the opening-flow relationship is a non-linear mapping). Pressure constraints are verified: the estimated outlet pressure (14.2 MPa) and deaerator pressure (0.749 MPa) at an opening of 58% are calculated and both are within limits.

[0067] Command fine-tuning and output: Overlay anti-cavitation verification: Ensure that the pressure after valve throttling (14.2MPa) is higher than the saturation pressure by 1.1MPa + a safety margin of 0.3MPa, and the condition is met. The final generated optimal valve opening command is 58%, with the accuracy rounded to 1%.

[0068] Signal Transmission: The opening command (58%) is sent to the electric actuator of the feedwater pump regulating valve via hard-wired connection (4-20mA analog signal) or bus (Profibus). The actuator has a built-in position feedback device (potentiometer or encoder) to detect the valve stem displacement in real time. Execution Verification: The actuator converts the electrical signal into valve disc stroke (corresponding to a stroke of 25mm) according to the equal percentage characteristic curve. If the feedback opening deviates from the command by more than 2% (feedback value <56% or >60%), an alarm is triggered and the system switches to standby control mode (fixed opening of 50%). Synergistic Effect: After the valve opening is adjusted, the feedwater flow rate drops to around 642t / h (close to the target value), and the deaerator pressure stabilizes at 0.749MPa, achieving synergistic suppression of minimum flow protection and pressure fluctuations.

[0069] In this embodiment of the invention, by deploying various types of sensors at different locations in the water supply system, key real-time operating parameters such as water supply flow rate, pump outlet pressure, pump speed, and deaerator pressure are comprehensively collected, providing a rich data foundation for precise system control. Fractional calculus is introduced into the description of the dynamic characteristics of the water supply system, enabling more accurate capture of the memory and historical dependencies of parameter changes, and better reflecting the actual dynamics of the system compared to traditional models. A multi-step prediction sequence containing predicted water supply flow rate and deaerator pressure values ​​can be generated, and the model parameters are dynamically fine-tuned by comparing them with real-time data through a residual detection mechanism, ensuring the reliability and adaptability of the prediction results. The multi-step prediction sequence is compared with preset values ​​to generate a collaborative error vector containing water supply flow rate tracking error and deaerator pressure tracking error, comprehensively reflecting system operating deviations and providing a comprehensive basis for subsequent control. Multiple constraints, such as valve characteristic constraints, flow safety constraints, pressure constraints, and anti-cavitation constraints, are set according to the unit's safe operation standards to ensure the system operates within a safe range. Opening commands are sent to the water pump regulating valve actuator, and the actuator's built-in position feedback device detects the valve stem displacement in real time and performs execution verification to ensure accurate command execution and improve system reliability.

[0070] Please see Figure 2-3 Another embodiment of the turbine feedwater pump minimum flow control system of the present invention includes:

[0071] 201. Acquisition module, used to acquire real-time operating parameters of the water supply system through sensors, and generate a set of real-time operating parameters including water supply flow rate, pump outlet pressure, pump speed and deaerator pressure;

[0072] Specifically, four raw sensor signals are generated: a flow sensor installed on the feedwater pump outlet pipe, a pressure sensor installed at the feedwater pump outlet, a speed sensor installed on the feedwater pump shaft, and a pressure transmitter installed on the top of the deaerator. These four raw sensor signals are then conditioned through filtering, amplification, and analog-to-digital conversion to generate four standardized digital signals. These standardized digital signals are input into a dynamic compensator to compensate for signal transmission delays based on the dynamic characteristics of the feedwater system, generating four real-time compensation signals. These four real-time compensation signals are then categorized and combined according to the parameter types of feedwater flow, pump outlet pressure, pump speed, and deaerator pressure to generate a parameter classification set. Each parameter in the parameter classification set is then dimensionless, and by dividing each parameter by its corresponding reference value, a real-time operating parameter set containing dimensionless feedwater flow, dimensionless pump outlet pressure, dimensionless pump speed, and dimensionless deaerator pressure is generated.

[0073] It should be noted that this example takes a 600MW thermal power unit participating in deep peak shaving, ensuring the minimum flow rate of the feedwater pump.

[0074] Sensors are installed at key process points to convert physical quantities into raw electrical signals. Flow acquisition: A vortex flow meter (flow sensor) is installed on the outlet pipe of the feedwater pump. When the feedwater flow rate is 320 tons / hour, a pulse signal with a frequency of 1250Hz is output. Pressure and speed acquisition: A pressure sensor installed at the feedwater pump outlet converts the 12.5MPa pressure into a 4-20mA current signal; a magnetoelectric speed sensor installed on the pump shaft converts the 5100rpm speed into a pulse signal; a pressure transmitter at the top of the deaerator converts the 0.8MPa pressure into a corresponding 4-20mA current signal. These raw analog and pulse signals enter the signal conditioning unit. For the current signal, it undergoes low-pass filtering (cutoff frequency set to 50Hz) to remove high-frequency noise caused by pump vibration, and then the analog input module performs analog-to-digital conversion (16-bit resolution) to convert the current value into a digital quantity between 0-65535. For the pulse signal, it is acquired by a high-speed counting module.

[0075] After conditioning, the digital signal still needs dynamic compensation and standardization to eliminate measurement lag and facilitate controller use. Dynamic compensation: Since sensors and transmission lines introduce measurement delays, these need to be corrected by a dynamic compensator. If the lag time constant of the pressure signal relative to the flow rate is known to be approximately 2 seconds, the compensator will use a lead correction network to predict the pressure change trend, making the signal closer to the actual operating conditions.

[0076] Dimensionless Conversion: To facilitate control algorithm processing, parameters with different dimensions need to be converted into dimensionless values. This is achieved by dividing each parameter by its corresponding reference value. The reference value for feedwater flow rate is set as the boiler's maximum continuous evaporation rate (approximately 690 t / h under BMCR conditions), the pump outlet pressure reference value is 15 MPa, the speed reference value is 5500 rpm, and the deaerator pressure reference value is 1.0 MPa. Therefore, dividing the currently measured flow rate of 320 t / h by the reference value of 690 t / h yields a dimensionless feedwater flow rate of approximately 0.46.

[0077] Generate a standardized set of real-time operating parameters: {dimensionless feedwater flow rate: 0.46, dimensionless pump outlet pressure: 0.83, dimensionless pump speed: 0.93, dimensionless deaerator pressure: 0.80}. This set accurately reflects the current state of the system.

[0078] 202. Prediction module, used to input the set of real-time operating parameters into a preset fractional-order state-space model to generate a multi-step prediction sequence containing predicted values ​​of feedwater flow rate and deaerator pressure.

[0079] Specifically, the real-time operating parameter set is fused with the historical state data sequence to generate an extended state vector containing current and historical state information. The extended state vector is input into a fractional-order state transition matrix to calculate the state evolution sequence for multiple future sampling times. The state evolution sequence is processed by output extraction to separate the feedwater flow evolution component and the deaerator pressure evolution component, generating an uncorrected prediction sequence. Based on the system disturbance characteristics and model error statistical features, the uncorrected prediction sequence is dynamically error corrected to generate a corrected prediction sequence. The corrected prediction sequence is truncated and recombined according to the prediction time domain length to generate a multi-step prediction sequence containing predicted feedwater flow and predicted deaerator pressure values.

[0080] Furthermore, the extended state vector is input into the fractional-order state transition matrix to calculate the state evolution sequence at multiple future sampling times. This includes: extracting typical operating modes of the water supply system under different load conditions from the historical operation database to generate a historical operation mode library; performing similarity matching between the extended state vector and the typical operating modes in the historical operation mode library to identify the feature parameters of the current operating mode and generate a mode matching result; based on the mode matching result, selecting the corresponding fractional-order differential operator from a preset fractional-order operator library to generate fractional-order calculation parameters adapted to the current operating state; using the fractional-order calculation parameters to perform fractional-order state transition calculations on the extended state vector to obtain the evolution trajectory of the system state at multiple future sampling times and generate an initial state evolution sequence; and combining the disturbance feature parameters acquired in real time to perform disturbance compensation on the initial state evolution sequence to generate a refined state evolution sequence.

[0081] It should be noted that during the deep peak shaving process of a 600MW thermal power unit (the load rapidly decreases from 450MW to 300MW), it is essential to ensure that the feedwater pump flow rate remains above its minimum safe value (210t / h). The following is a specific example of how this step is implemented:

[0082] By integrating real-time parameters (dimensionless feedwater flow rate 0.72, deaerator pressure 0.75) with historical state sequences (flow rate, pressure, and rotational speed data for the past 30 seconds), an extended state vector containing current and historical information is generated. The vector is represented as: [current flow rate, flow rate of the previous 1 second, flow rate of the previous 2 seconds, current pressure, pressure of the previous 1 second, rotational speed...].

[0083] Pattern Matching and Fractional Parameter Selection: The system performs similarity matching between the current extended state vector and the historical operating mode library ("300MW low load steady state" and "load rapid drop condition"). If the current state matches the "load rapid drop" mode by 85%, a corresponding fractional differential operator (order α=0.7) is selected from the fractional operator library to adapt to the dynamic process. State Evolution Calculation: The system uses fractional parameters to perform state transition calculations on the extended state vector, generating an initial state evolution sequence for the next 30 seconds (one point every 5 seconds). The predicted feedwater flow rate gradually decreases from the current 0.72 to 0.68, and the deaerator pressure increases from 0.75 to 0.78. Disturbance Compensation: Based on the real-time collected disturbance characteristics (sudden drop in boiler combustion rate of 10t / h), the initial sequence is compensated. Due to the decrease in combustion rate causing the evaporation section to shift later, the predicted flow rate needs to be corrected upwards by 0.02, generating a refined state evolution sequence.

[0084] Based on model error statistics (historical prediction error standard deviation is 0.015), the uncorrected prediction sequence is calibrated, and a deviation correction of -0.01 is added to the flow prediction value. Finally, the multi-step prediction sequence of the next 5 sampling points is output: water flow prediction value: [0.71, 0.69, 0.68, 0.67, 0.66] (dimensionless); deaerator pressure prediction value: [0.76, 0.77, 0.78, 0.78, 0.79] (dimensionless).

[0085] 203. Processing module, used to compare the multi-step prediction sequence with the preset feedwater flow rate setpoint and deaerator pressure setpoint, and generate a collaborative error vector including feedwater flow rate tracking error and deaerator pressure tracking error;

[0086] Specifically, based on the current unit load parameters and deaerator operating conditions, the feedwater flow rate reference setpoint and deaerator pressure reference setpoint are extracted from the preset load-pressure characteristic curve to generate a setpoint sequence for multiple future sampling times. The predicted feedwater flow rate in the multi-step prediction sequence is compared point-by-point with the feedwater flow rate reference setpoint in the setpoint sequence, and the predicted deaerator pressure is also compared point-by-point with the deaerator pressure reference setpoint to generate a preliminary error sequence. Dynamic coupling analysis is performed on the preliminary error sequence to identify the coupling relationship between feedwater flow rate error and deaerator pressure error, generating an error coupling feature matrix. Based on the error coupling feature matrix, the preliminary error sequence is weighted and reconstructed to generate a dynamic error sequence containing time-related characteristics. The dynamic error sequence is then normalized and time-series integrated to generate a collaborative error vector containing feedwater flow rate tracking error and deaerator pressure tracking error.

[0087] Furthermore, dynamic coupling analysis is performed on the preliminary error sequence to identify the coupling relationship between feedwater flow error and deaerator pressure error, generating an error coupling feature matrix. This includes: extracting dynamic response characteristic parameters of the feedwater system under different operating conditions from the system characteristic database to generate a system dynamic coupling model; analyzing the influence characteristics of feedwater flow rate changes on deaerator pressure and deaerator pressure changes on feedwater flow rate based on the system dynamic coupling model to generate a two-way coupling influence factor; fusing the two-way coupling influence factor with the preliminary error sequence to obtain the dynamic coupling strength between errors and generating a coupling strength distribution map; establishing an error propagation path model based on the coupling strength distribution map, identifying the dominant and secondary error propagation paths, and generating an error propagation network; and constructing a feature matrix reflecting the coupling relationship between feedwater flow rate error and deaerator pressure error based on the error propagation network to generate the error coupling feature matrix.

[0088] It should be noted that, in the scenario of deep peak shaving of a 600MW unit to a 300MW load, the implementation example of step 203 to ensure the minimum flow rate of the feedwater pump (safety threshold 210t / h, dimensionless 0.65) is as follows:

[0089] Based on the current unit load of 300MW (decreasing at a rate of 1MW / 5s) and the deaerator operating conditions, the setpoint sequence for the next 5 sampling times (5-second intervals) is extracted from the preset load-pressure characteristic curve: Feedwater flow rate setpoint: linearly decreasing from the baseline value of 300t / h (dimensionless 0.435) according to the load decrease trend. Deaerator pressure setpoint: synchronously fine-tuned from the baseline value of 0.7MPa (dimensionless 0.7).

[0090] The setpoint sequence is shown in the table below:

[0091]

[0092] A preliminary error sequence is generated from the multi-step prediction sequence in step 202: predicted water flow rate: [0.72, 0.69, 0.67, 0.65, 0.63] (dimensionless); predicted deaerator pressure: [0.76, 0.77, 0.78, 0.79, 0.80] (dimensionless).

[0093] After point-by-point comparison, a preliminary error sequence is generated: Water flow error: [0.285, 0.256, 0.237, 0.218, 0.199] (positive error indicates that the flow rate is too high); Deaerator pressure error: [0.060, 0.071, 0.082, 0.093, 0.104] (positive error indicates that the pressure is too high);

[0094] Extracting dynamic response characteristics from the system characteristic database: Two-way coupling influence factor: Influence coefficient of feedwater flow rate change on deaerator pressure. =3.45 (A 0.0145 dimensionless unit increase in flow rate leads to a 0.05 dimensionless unit increase in pressure); the coefficient of influence of deaerator pressure change on feedwater flow rate. =0.0725 (A pressure increase of 0.1 dimensionless unit results in a flow rate decrease of 0.00725 dimensionless units).

[0095] Coupling strength calculation formula: ;

[0096] in This indicates the coupling strength between flow rate error and pressure error. Conversely. Taking point t+5s as an example:

[0097]

[0098]

[0099] The coupling strength distribution shows that the dominant path is the influence of flow error on pressure error (Sfp is much larger than Spf). Construct the error coupling characteristic matrix M:

[0100]

[0101] Weighted reconstruction: The error vector at each time step is weighted using matrix M. For example, at time t+5s:

[0102] .

[0103] Normalization: The dynamic error sequence is scaled to the range [0,1] by dividing by the maximum value (maximum flow error 0.492, maximum pressure error 0.104). Time series integration: The average value of 5 time points is taken to generate a collaborative error vector [0.217, 0.082], representing the combined tracking errors of feedwater flow and deaerator pressure, respectively.

[0104] 204. Setting module, used to input the cooperative error vector into the adaptive fractional sliding surface for processing, to generate a sliding control quantity that simultaneously includes the feedwater flow control component and the deaerator pressure stability component;

[0105] Specifically, based on the current operating conditions of the system, fractional-order parameters, convergence rate parameters, and coupling weight parameters are obtained from the preset sliding mode parameter configuration table to generate the initial parameter configuration of the sliding mode surface. The cooperative error vector is input into the fractional-order differential processor to calculate the fractional-order differential sequences of the feedwater flow tracking error and the deaerator pressure tracking error, respectively, generating an error differential feature set. Based on the error differential feature set and the initial parameter configuration of the sliding mode surface, the basic sliding mode surface quantity is calculated through the sliding mode surface dynamic equation. According to the system state change rate, the initial parameter configuration of the sliding mode surface is adaptively adjusted online to generate real-time optimized sliding mode surface parameters. The basic sliding mode surface quantity is reconstructed using the real-time optimized sliding mode surface parameters to generate a sliding mode control quantity that simultaneously includes the feedwater flow control component and the deaerator pressure stability component.

[0106] Furthermore, based on the system state change rate, the initial parameter configuration of the sliding surface is adaptively adjusted online to generate real-time optimized sliding surface parameters. This includes: real-time monitoring of the state change characteristics of the water supply system, including the rate of change of water flow, the rate of change of deaerator pressure, and the operating frequency of regulating valves, to generate a system state change rate feature set; based on the system state change rate feature set, analyzing the matching relationship between the system dynamic response characteristics and the sliding surface parameters, and generating a parameter adjustment strategy; according to the parameter adjustment strategy, coordinating the adjustment of fractional-order parameters, convergence rate parameters, and coupling weight parameters to generate initially adjusted sliding surface parameters; evaluating the stability margin of the initially adjusted sliding surface parameters through a stability verification mechanism to generate parameter stability verification results; and based on the parameter stability verification results, performing final optimization of the initially adjusted sliding surface parameters to generate real-time optimized sliding surface parameters.

[0107] It should be noted that this applies to scenarios where a 600MW unit undergoes deep peak shaving to a 300MW load, while ensuring minimum feedwater pump flow:

[0108] The cooperative error vector is the water flow tracking error e. f =0.217 and deaerator pressure tracking error e p =0.082. Based on the current "rapid load drop" condition, the control system calls the initial parameters from the preset sliding mode parameter configuration table: fractional order α=0.7, convergence rate parameter λ=0.5, coupling weight w=0.6 (the water flow control weight is higher than the pressure stability weight).

[0109] The cooperative error vector is fed into a fractional-order differential processor. The fractional-order differential calculation considers not only the current value of the error but also its historical variation. The fractional-order differential sequence for the feedwater flow tracking error is calculated as [0.217, 0.201, 0.189, ...], and the fractional-order differential sequence for the deaerator pressure tracking error is [0.082, 0.079, 0.075, ...]. These two sequences together constitute the error differential feature set.

[0110] Based on the error differential feature set and initial parameters, the basic sliding surface quantity s0 is calculated using the sliding surface dynamic equation:

[0111]

[0112] in This represents a fractional differential operator. Substituting specific numerical values ​​into the estimation, we set... We get a base value of 0.35.

[0113] The system's real-time monitoring of state change rates revealed: feedwater flow rate change rate was -12 t / h / min, deaerator pressure change rate was +0.05 MPa / min, and the regulating valve actuation frequency was high (5 times in the last 30 seconds). Analysis indicated that the system was in a rapid dynamic process, requiring stronger control. Therefore, the parameter adjustment strategy was to increase the convergence rate parameter λ from 0.5 to 0.8 to accelerate the response, while fine-tuning the coupling weight w from 0.6 to 0.55 to prevent over-correction of pressure errors from affecting flow-dominant control. The fractional order α was maintained at 0.7 to preserve the system's historical state memory characteristics. A stability verification mechanism assessed that the system had sufficient stability margin (phase margin greater than 45 degrees) under this parameter combination, confirming the parameters' usability.

[0114] Using real-time optimized sliding surface parameters (α=0.7, λ=0.8, w=0.55), the basic sliding surface quantity s0 was adjusted.

[0115] The system is reconstructed to generate the final sliding mode control variable u. This control variable is a comprehensive command with a calculated value of 0.61. This value incorporates both the feedwater flow control component (dominant, approximately 65%) and the deaerator pressure stabilization component (auxiliary, approximately 35%), aiming to synergistically eliminate errors and ensure that the feedwater pump flow rate remains above the minimum safe value during rapid load reduction, while suppressing excessive fluctuations in deaerator pressure.

[0116] 205. The allocation module is used to solve for the optimal opening command of the regulating valve based on the sliding mode control quantity and the preset system operation constraints, and send the command to the feedwater pump regulating valve actuator to realize the coordinated control of the minimum flow rate of the feedwater pump and the deaerator pressure.

[0117] Specifically, the mechanical limit constraints, action rate constraints, and safe operation constraints of the feedwater pump regulating valve are extracted from the system configuration database to generate a multi-dimensional system constraint set. The sliding mode control quantity is input into the constraint optimization processor, and combined with the multi-dimensional system constraint set, iterative solutions are obtained through a multi-objective optimization algorithm to generate a preliminary optimized opening sequence. The preliminary optimized opening sequence is dynamically smoothed to eliminate abrupt changes in control commands, generating a smooth opening command sequence. This smooth opening command sequence is input into the valve characteristic compensator, and feedforward compensation is performed based on the nonlinear flow characteristics of the regulating valve to generate a compensated opening command sequence. The optimal opening command for the current moment is extracted from the compensated opening command sequence, converted into an analog control signal via a digital-to-analog converter, and sent to the feedwater pump regulating valve actuator to achieve coordinated control of the feedwater pump minimum flow rate and deaerator pressure.

[0118] It should be noted that the theoretical sliding mode control value (0.61) calculated in the previous step is transformed into an optimal opening command that can be directly executed under various safety and mechanical constraints and can effectively control the feedwater pump recirculation regulating valve.

[0119] The control system retrieves a pre-defined set of multi-dimensional system constraints from the configuration database. The main contents of this set are shown in the table below.

[0120]

[0121] The sliding mode control variable is fed into the constraint optimization processor for iterative solution: Multi-objective optimization solution: Based on the intention of "prioritizing flow rate while considering pressure stability" implied by the sliding mode control variable (0.61), the optimization algorithm solves the problem under the premise of satisfying all the above constraints. Its goal is to find a regulating valve opening that can quickly increase the feedwater flow rate above the safe threshold while avoiding drastic pressure fluctuations in the deaerator and excessive valve action. The initial solution yields an opening sequence [42%, 45%, 48%]. Dynamic smoothing: To prevent frequent valve action, the initial optimized opening sequence is smoothed by filtering to eliminate abrupt changes, smoothing the sequence to [43%, 45%, 47%], making the command changes more uniform. Valve characteristic compensation: Feedwater pump recirculation regulating valves usually have equal percentage flow characteristics, that is, the flow rate changes slowly at small openings and drastically at large openings. To ensure linear control, the valve characteristic compensator will perform feedforward compensation based on this characteristic. To achieve the theoretical flow rate, the command needs to be further modified to [46%, 48%, 50%].

[0122] The optimal opening command (46%) for the current moment is extracted from the compensated opening command sequence. This digital command is converted into a 4-20mA analog current signal by a digital-to-analog converter (D / A) and sent to the electro-pneumatic valve positioner of the feedwater pump recirculation regulating valve.

[0123] After receiving the signal, the positioner drives the actuator to precisely open the valve to 46%. At this time, a portion of the feedwater returns to the deaerator through the recirculation pipeline, thereby ensuring that the flow rate through the feedwater pump body is higher than its minimum safe value, effectively preventing cavitation inside the pump, and minimizing interference with the deaerator pressure, thus achieving the goal of coordinated control.

[0124] 206. Protection Module: This module monitors key operating parameters of the feedwater pump system in real time, including instantaneous feedwater flow rate, deaerator pressure, and regulating valve opening, generating a real-time system status monitoring set. It compares this set with preset safe operating thresholds to identify abnormal deviations in system operating status and generates a system safety status assessment result. Based on this assessment, it dynamically adjusts the protective constraint boundaries of the control system, generating adaptive safety constraint boundaries. When the system operating status approaches the safety boundary, a preventative control mechanism is activated, generating protective control compensation quantities. These compensation quantities are integrated into the main control loop to enhance the system's safety margin without interrupting normal control, achieving enhanced safety operation of the feedwater pump minimum flow control system. The protective control compensation quantities are ultimately integrated into the main control loop to achieve enhanced safety control.

[0125] It should be noted that the system continuously monitors key parameters and compares them with preset safety thresholds in real time to generate a safety status assessment result.

[0126]

[0127] Based on the table above, the system identified that the feedwater flow rate (208 t / h) was below the minimum safe flow rate (210 t / h), and the deaerator pressure (0.48 MPa) was also close to the lower operating limit (0.5 MPa). The control system assessment concluded that the system's operating state was close to the safety boundary, and there was a risk of feedwater pump cavitation and excessively low deaerator pressure.

[0128] Based on the above evaluation results, the system dynamically adjusts its internal control constraint boundaries to achieve preventative protection. Flow command safety boundary: The upper limit of the water supply flow control command is tightened from 300t / h to 250t / h to prevent excessive command during flow increases from causing abrupt valve action and further deterioration of system pressure. Valve opening change rate constraint: The maximum allowable opening change rate of the control valve is reduced from 2% per second to 0.5% per second, aiming to smooth control actions and avoid drastic fluctuations in water flow and pressure.

[0129] When the operating point is detected to be continuously approaching the safety boundary, the preventive control mechanism is triggered. Protective control compensation generation: The control algorithm calculates the compensation required to escape the current dangerous state, generating an additional feedwater flow demand compensation value of +15t / h, and generating a deaerator pressure stabilization feedforward compensation signal, which slightly increases the deaerator pressure setpoint.

[0130] Compensation Quantity Integration and Safety Enhancement: The aforementioned protective control compensation quantity is seamlessly integrated into the main control loop (sliding mode control quantity in step 204). When calculating the final regulating valve opening command, the controller prioritizes meeting the safety requirements represented by this compensation quantity. This increases the feedwater flow command by 15 t / h, thereby guiding the regulating valve opening to increase smoothly within the constraint range, helping the flow rate return to above the safe range as quickly as possible, while simultaneously suppressing further drops in deaerator pressure.

[0131] In this embodiment of the invention, a set of real-time operating parameters is input into a preset fractional-order state-space model to generate a multi-step prediction sequence containing predicted values ​​for feedwater flow and deaerator pressure. This allows for advance understanding of future system state changes and provides a forward-looking basis for control decisions. The real-time parameters are fused with historical state data sequences to generate an extended state vector. This vector is then used for pattern matching with a historical operating mode library, and a suitable fractional-order differential operator is selected to make the predictions more consistent with the actual operating characteristics of the system, thus improving prediction accuracy. Based on system disturbance characteristics and model error statistical features, dynamic error correction is performed on the uncorrected prediction sequence to further optimize the prediction results and enhance the system's ability to cope with complex operating conditions. Finally, the multi-step prediction sequence is compared with setpoints to generate a collaborative error vector containing feedwater flow tracking errors and deaerator pressure tracking errors, comprehensively reflecting system control errors. In poor conditions, dynamic coupling analysis is used to identify the coupling relationship between feedwater flow error and deaerator pressure error, generating an error coupling feature matrix to provide a basis for accurately handling error relationships and avoid control errors caused by error coupling. The initial optimized opening sequence is dynamically smoothed to eliminate abrupt changes in control commands, and then feedforward compensation is performed based on the nonlinear flow characteristics of the regulating valve to generate a compensated opening command sequence, making the valve action smooth and the control effect linear, thus improving the system control quality. When the system operating state is detected to be close to the safety boundary, a preventive control mechanism is activated to generate protective control compensation and integrate it into the main control loop. This enhances the system safety margin without interrupting the normal control process, realizing the safe enhanced operation of the feedwater pump minimum flow control system and effectively preventing problems such as feedwater pump cavitation and deaerator pressure abnormalities.

[0132] Figure 4This is a schematic diagram of a minimum flow control device for a steam turbine feedwater pump according to an embodiment of the present invention. The minimum flow control device 300 for the steam turbine feedwater pump may include a processor 301 and a memory 302. The memory 302 is used to store program instructions and / or data, and the processor 301 is used to execute the program instructions stored in the memory 302, thereby implementing the method described in the above-described method embodiment.

[0133] Optionally, the memory 302 and the processor 301 are coupled. The coupling is an indirect coupling or communication connection between devices, units, or modules, and can be electrical, mechanical, or other forms, for information interaction between devices, units, or modules.

[0134] Optionally, the turbine feedwater pump minimum flow control device 300 may further include a communication interface 303. The communication interface 303 is used to communicate with other devices via a transmission medium, for example, transmitting received signals from other communication devices to the processor 301, or transmitting signals from the processor 301 to other communication devices. The communication interface 303 may be a transceiver or an interface circuit, such as a transceiver circuit or a transceiver chip.

[0135] This application embodiment does not limit the specific connection medium between the processor 301, memory 302, and communication interface 303. This application embodiment... Figure 4 The processor 301, memory 302, and communication interface 303 are connected via a bus 304. Figure 4 The connections between other components are shown in bold and are for illustrative purposes only, not as limiting information. The bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, Figure 4 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0136] The present invention also provides a minimum flow control device for a steam turbine feedwater pump. The steam turbine feedwater pump minimum flow control device includes a memory and a processor. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor performs the steps of the minimum flow control system for the steam turbine feedwater pump in the above embodiments.

[0137] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the turbine feedwater pump minimum flow control system.

[0138] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0139] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0140] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A minimum flow control system for a steam turbine feedwater pump, characterized in that, include: The acquisition module is used to acquire real-time operating parameters of the water supply system and generate a real-time operating parameter set, including: fusing the real-time operating parameter set with historical state data sequences to generate an extended state vector; inputting the extended state vector into a fractional-order state transition matrix to calculate a state evolution sequence for multiple future sampling times; performing output extraction processing based on the state evolution sequence to separate the water supply flow evolution component and the deaerator pressure evolution component, generating an uncorrected prediction sequence; performing dynamic error correction on the uncorrected prediction sequence to generate a corrected prediction sequence; and truncating and recombining the corrected prediction sequence according to the prediction time domain length to generate a multi-step prediction sequence. The prediction module is used to input the set of real-time operating parameters into a preset fractional-order state-space model to generate a multi-step prediction sequence. The processing module is used to compare the multi-step prediction sequence with preset feedwater flow rate setpoints and deaerator pressure setpoints to generate a collaborative error vector. This includes: extracting the feedwater flow rate reference setpoint and deaerator pressure reference setpoint from a preset load-pressure characteristic curve based on the current unit load parameters and deaerator operating conditions, generating a setpoint sequence for multiple future sampling times; comparing the predicted feedwater flow rate in the multi-step prediction sequence with the feedwater flow rate reference setpoint in the setpoint sequence point-by-point, and simultaneously comparing the predicted deaerator pressure with the deaerator pressure reference setpoint point-by-point, generating a preliminary error sequence; performing dynamic coupling analysis based on the preliminary error sequence to identify the coupling relationship between feedwater flow rate error and deaerator pressure error, generating an error coupling feature matrix; weighting and reconstructing the preliminary error sequence based on the error coupling feature matrix to generate a dynamic error sequence; and processing the dynamic error sequence to generate a collaborative error vector. The configuration module is used to process the cooperative error vector input into an adaptive fractional sliding surface to generate sliding mode control variables. This includes: obtaining fractional-order parameters, convergence rate parameters, and coupling weight parameters from a preset sliding mode parameter configuration table based on the current system operating conditions, and generating an initial sliding surface parameter configuration; calculating the fractional-order differential sequences of the feedwater flow tracking error and the deaerator pressure tracking error based on the cooperative error vector, and generating an error differential feature set; calculating the basic sliding surface quantity using the sliding surface dynamic equation based on the error differential feature set and the initial sliding surface parameter configuration; adaptively adjusting the initial sliding surface parameter configuration online according to the system state change rate to generate real-time optimized sliding surface parameters; and reconstructing the basic sliding surface quantity using the real-time optimized sliding surface parameters to generate sliding mode control variables. The allocation module is used to solve for the optimal opening command of the regulating valve based on the sliding mode control quantity and the preset system operation constraints.

2. The turbine feedwater pump minimum flow control system according to claim 1, characterized in that, include: Extract typical operating modes of the water supply system under different load conditions from the historical operation database to generate a historical operation mode library; The extended state vector is matched with typical operating modes in the historical operating mode library to identify the feature parameters of the current operating mode and generate a pattern matching result. Based on the pattern matching results, a corresponding fractional differential operator is selected from a preset fractional operator library to generate fractional calculation parameters that are adapted to the current running state. The fractional-order calculation parameters are used to perform fractional-order state transition calculations on the extended state vector to obtain the evolution trajectory of the system state at multiple future sampling times, thereby generating the initial state evolution sequence. The initial state evolution sequence is perturbed by combining the perturbation feature parameters collected in real time, and a refined state evolution sequence is generated.

3. The turbine feedwater pump minimum flow control system according to claim 1, characterized in that, Based on the error differential feature set and initial parameters, the basic sliding surface quantity s0 is calculated: in, This is to account for the error in water supply flow rate; This refers to the deaerator pressure error. λ is a fractional differential operator; λ is the convergence rate parameter; w is the coupling weight.

4. The turbine feedwater pump minimum flow control system according to claim 1, characterized in that, include: Real-time monitoring of the state change characteristics of the water supply system generates a system state change rate feature set; Based on the system state change rate feature set, the matching relationship between the system dynamic response characteristics and the sliding surface parameters is analyzed, and a parameter adjustment strategy is generated. According to the parameter adjustment strategy, the fractional order parameter, convergence rate parameter and coupling weight parameter are adjusted in a coordinated manner to generate the preliminary adjusted sliding surface parameters. The stability margin of the initially adjusted sliding surface parameters is evaluated through a stability verification mechanism, and parameter stability verification results are generated. Based on the stability verification results of the parameters, the initially adjusted sliding surface parameters are finally optimized to generate real-time optimized sliding surface parameters.

5. The turbine feedwater pump minimum flow control system according to claim 1, characterized in that, include: Extract the mechanical limit constraints, action rate constraints, and safe operation constraints of the water supply system from the water supply pump regulating valve to generate a multi-dimensional system constraint set; Based on the sliding mode control quantity and the multi-dimensional system constraint set, an iterative solution is performed using a multi-objective optimization algorithm to generate a preliminary optimized opening sequence; The preliminary optimized opening sequence is dynamically smoothed to eliminate abrupt changes in control commands and generate a smooth opening command sequence. Based on the smooth opening command sequence, feedforward compensation is performed based on the nonlinear flow characteristics of the control valve to generate a compensated opening command sequence. The optimal opening command for the current moment is extracted from the compensated opening command sequence, converted into an analog control signal by a digital-to-analog converter, and sent to the feedwater pump regulating valve actuator to achieve coordinated control of the feedwater pump minimum flow and deaerator pressure.

6. The turbine feedwater pump minimum flow control system according to claim 1, characterized in that, It also includes a protection module: Real-time monitoring of key operating parameters of the water supply pump system generates a real-time system status monitoring set; The system's real-time status monitoring set is compared with a preset safe operation threshold to identify the degree of abnormal deviation in the system's operating status and generate a system safety status assessment result. Based on the system safety status assessment results, the protective constraint boundaries of the control system are dynamically adjusted to generate adaptive safety constraint boundaries; When the system's operating state is detected to be approaching the safety boundary, a preventive control mechanism is activated to generate protective control compensation. The protective control compensation is integrated into the main control loop to enhance the system's safety margin without interrupting the normal control process, thereby achieving enhanced safe operation of the feedwater pump minimum flow control system.

Images