Quick start-stop control system and method for steam turbine

Abstract

The invention relates to the technical field of control adjustment, in particular to a steam turbine quick start-stop control system and method, which comprises a high-precision sensing module, a thermal stress reconstruction module, a self-adaptive decision module, a vibration suppression execution module and a cooling optimization execution module. Acquiring data through an embedded optical fiber matrix which is arranged in an axial and radial crossed manner and a curved surface fitting infrared array, and fusing metal phase change parameters to reconstruct a rasterized stress nephogram; the adaptive decision-making module analyzes the stress distribution characteristics, and generates a control signal according to a thermal state risk level switching calculation model; the vibration suppression execution module fuses the vibration mode and the temperature adjusting quantity to dynamically distribute a vibration suppression strategy; and the digital twin module for closed-loop verification calibrates model parameters through crack prediction and actually measured data deviation, so that the reliability of the system is improved. The high-pressure rotor thermal stress field reconstruction distortion problem is solved, the crack initiation risk is reduced, and the service life of equipment is prolonged.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of control adjustment technology, in particular to a steam turbine fast start-stop control system and method. BACKGROUND

[0002] With the aggravation of power load changes brought by the popularization of renewable energy, the power grid needs to frequently adjust the power generation power to achieve supply and demand balance, and requires the steam turbine to have fast start-stop capability to adapt to dynamic demand changes. The conventional start-stop process is limited by the stress accumulation caused by uneven thermal expansion of metal parts, which may cause equipment failure or prolong unplanned downtime, reducing overall efficiency; therefore, the fast start-stop control uses high-precision sensors to monitor the temperature gradient evolution of key parts, and combines with automatic algorithm to adjust the steam parameters in real time, minimizes the stress peak by optimizing the heating and cooling rate, and shortens the start-stop cycle.

[0003] The core pain point of the steam turbine fast start-stop control system is the feedback delay between the real-time coordination of the thermal stress minimization of complex geometric components and the control decision. Due to the multi-surface characteristics of structures such as the root of the high-pressure rotor impeller, the temperature field reconstruction accuracy is insufficient, and the finite element inversion algorithm relies on discrete sensor data to calculate the internal stress distribution. However, the dynamic changes of steam parameters in the actual start-stop process will introduce measurement errors, such as the sudden mutation of local thermal gradient when high-temperature steam impacts the impeller, which is not captured in time, causing stress prediction deviation, triggering a conservative control strategy to prolong the start-stop cycle or excessively pursuing speed leading to crack risk; for example, during the start-up stage, the intelligent curve planning expects uniform heating, and the fiber temperature measurement point spacing of a certain impeller root is too large to miss the local hot spot. The algorithm does not adjust the steam flow temperature in time, the instantaneous uneven stress exceeds the standard and causes micro-cracks, which need to be repaired by stopping the machine, weakening the power grid response efficiency and equipment reliability. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a steam turbine fast start-stop control system and method, which solves the problem of real-time reconstruction distortion of the temperature field caused by the thermal mechanical response lag of the core metal components of the high-pressure rotor in complex structures.

[0005] To solve the above technical problems, the specific technical solutions of the present application are as follows:

[0006] In a first aspect, the present application provides a steam turbine fast start-stop control system, comprising:

[0007] A high-precision sensing module is arranged at the root of the high-pressure rotor impeller, which collects the internal temperature gradient, surface transient temperature field and micro-deformation characteristics of the metal, and generates a three-dimensional temperature and strain tensor set with spatial coordinates;

[0008] a thermal stress reconstruction module receiving the three-dimensional temperature and strain tensor set, enhancing the resolution of the curved surface area through a dynamic grid encryption strategy, combining metal phase change parameters to reconstruct the temperature field and stress field in real time, and outputting a rasterized stress cloud map;

[0009] an adaptive decision-making module analyzing the stress accumulation rate and spatial distribution characteristics of the rasterized stress cloud map, switching calculation models according to the thermal state risk level, and generating control signals including steam valve opening degree instructions and temperature adjustment amounts;

[0010] a vibration suppression execution module receiving the temperature adjustment amount in the control signal, identifying the rotor vibration modal characteristics in real time, matching the shaft system resonance speed prediction result to dynamically generate a hydraulic damping force field and a throttle flow distribution strategy;

[0011] a cooling optimization execution module receiving the steam valve opening degree instruction in the control signal, encoding the steam pressure change trajectory, steam temperature gradient curve and drain valve action sequence into a chromosome structure during the shutdown stage, and optimizing the cooling path through a multi-objective evolutionary algorithm;

[0012] wherein the output of the high-precision sensing module drives the operation of the thermal stress reconstruction module;

[0013] The rasterized stress cloud map output by the thermal stress reconstruction module triggers the model scheduling of the adaptive decision-making module;

[0014] The adaptive decision-making module transmits control signals by channel: temperature adjustment amount to the vibration suppression execution module, and steam valve opening degree instruction to the cooling optimization execution module.

[0015] Further, the steam turbine rapid start-stop control system of the present application, the high-precision sensing module comprises:

[0016] The embedded optical fiber matrix is arranged in the blade root groove area in an axial and radial cross layout mode;

[0017] The curved surface fitted infrared array covers the groove surface area and compensates for motion blur errors;

[0018] The collection data of the embedded optical fiber matrix and the curved surface fitted infrared array are input to the space-time calibration engine;

[0019] The space-time calibration engine binds the physical coordinates of the sensors through a laser positioning system, synchronizes the time stamps of multiple source data, and outputs the calibrated sensor data set;

[0020] The calibrated sensor data set drives the interpolation algorithm to fuse and generate the three-dimensional temperature field.

[0021] Further, the steam turbine rapid start-stop control system of the present application, the thermal stress reconstruction module comprises:

[0022] a dynamic material property management subsystem storing austenite / ferrite phase transition temperature related Young's modulus parameters;

[0023] an intelligent mesh subdivision engine receiving the three-dimensional temperature and strain tensor set, performing mesh density dynamic adjustment according to real-time steam flow rate change rate, recursively subdividing blade root groove area mesh in high flow rate working conditions, and inserting transition layer mesh in curvature mutation areas;

[0024] an output of the intelligent mesh subdivision engine is connected to a thermal delay correction system, which loads the steam flow state mapping table to reconstruct rotor surface heat exchange boundary conditions and constrain inversion calculation iteration convergence speed;

[0025] updated boundary conditions output by the thermal delay correction system are input into a grid inversion model of the dynamic material property management subsystem to generate a gridded stress cloud map.

[0026] Further, the steam turbine rapid start-stop control system provided by the application comprises:

[0027] a dynamic model routing gateway receiving the gridded stress cloud map, extracting stress accumulation rate and spatial distribution characteristics, and performing calculation model switching according to the thermal state risk level: a simplified stress calculation model is called when an axial temperature standard deviation meets a first threshold value, and a high-frequency iteration model is activated when a groove area temperature rise rate exceeds a second threshold value;

[0028] stress evaluation results output by the dynamic model routing gateway are input into a spatial grid management subsystem, which discretizes the stress field into variable-size cubic cells, automatically encrypts grid distribution in high-stress areas and marks elastic domain or plastic domain double-threshold states, and generates a risk evaluation signal to be fed back to the dynamic model routing gateway;

[0029] the risk evaluation signal of the spatial grid management subsystem drives generation of a control signal.

[0030] Further, the steam turbine rapid start-stop control system provided by the application comprises:

[0031] a vibration sensing unit, which arranges a MEMS sensor cluster in a spatial spiral layout, collects rotor three-axis acceleration characteristics and generates original vibration signals;

[0032] a resonance pre-judgment engine receiving the original vibration signals for joint time-frequency domain analysis, and comparing historical vibration mode atlas library to identify bending or torsional mode characteristic vectors;

[0033] The output of the resonance prediction engine is connected to a vibration suppression strategy execution unit, which simultaneously receives the temperature adjustment amount in the control signal, dynamically allocates the phase angle of the hydraulic actuator according to the modal characteristic vector fusion temperature adjustment parameter, and adjusts the opening ratio of the DEH system high-pressure control valve group and the steam supplement valve.

[0034] Further, the steam turbine rapid start-stop control system provided by the application comprises a cooling optimization execution module.

[0035] A control instruction analysis unit receives the steam valve opening instruction, extracts the steam pressure change trajectory, the steam temperature gradient curve and the drain valve action sequence.

[0036] The output of the control instruction analysis unit is connected to a gene coding system, which encapsulates the extracted steam pressure change trajectory, steam temperature gradient curve and drain valve action sequence into a chromosome structure.

[0037] The chromosome structure of the gene coding system is input into a fitness evaluation system, which executes a multi-objective evolutionary algorithm calculation process to calculate the weighted evaluation values of the stress peak suppression rate, the superheat degree maintenance index and the shutdown time compression rate.

[0038] The output of the fitness evaluation system drives an evolutionary execution system to run, which adopts a time-sharing gene crossover strategy to recombine the control chain sequence and forcibly checks the main steam superheat degree boundary constraint in the iteration process, and outputs the optimized cooling path instruction.

[0039] Further, in the running process of the cooling optimization execution module of the steam turbine rapid start-stop control system provided by the application,

[0040] The fitness evaluation system acquires the rotor life state parameters in real time, and executes a dynamic weight distribution strategy according to the rotor life state parameters.

[0041] Under the working condition of a new unit, a high weight is distributed to the shutdown time compression rate.

[0042] Under the working condition of an aging unit, a high weight is distributed to the stress peak suppression rate.

[0043] The output of the weight distribution strategy is fed back to the evolutionary execution system in real time to drive the adjustment of the gene crossover strategy in the iteration process.

[0044] Further, the steam turbine rapid start-stop control system provided by the application further comprises a closed-loop verified digital twin module.

[0045] A data fusion gateway receives the gridded stress cloud map of the thermal stress reconstruction module and the vibration frequency spectrum data of the vibration suppression execution module.

[0046] A holographic data warehouse system connected to the output of the data fusion gateway stores rasterized stress snapshots and vibration frequency spectra in a four-dimensional spatiotemporal index;

[0047] A damage evolution prediction engine loads historical stress data from the holographic data warehouse system and calculates a crack propagation curve based on local strain amplitude;

[0048] A parameter calibration microservice receives the crack propagation curve and residual life prediction value from the damage evolution prediction engine and feeds a rotor residual life percentage to a threshold management system of the adaptive decision module.

[0049] Further, the parameter calibration microservice of the steam turbine rapid start-stop control system establishes a dynamic feedback channel:

[0050] Upon receiving the crack propagation curve and residual life prediction value from the damage evolution prediction engine,

[0051] When the residual life prediction value is lower than a set threshold,

[0052] A threshold contraction instruction is generated and transmitted to a spatial grid management subsystem of the adaptive decision module,

[0053] The elastic domain warning threshold is adjusted downward to a preset safety margin;

[0054] An endoscopic real-time data acquisition unit acquires a rotor surface crack distribution image and inputs it to the parameter calibration microservice;

[0055] The parameter calibration microservice performs crack position deviation analysis:

[0056] By comparing the crack distribution image with the coordinates of the predicted crack position, a spatial position offset is calculated, and a confidence correction coefficient is generated based on the offset;

[0057] The confidence correction coefficient is input to the damage evolution prediction engine to update the calculation parameters of the crack propagation curve.

[0058] In a second aspect, the present application provides a steam turbine rapid start-stop control method, which is applied to the steam turbine rapid start-stop control method and includes:

[0059] Step 1: Deploy a sensor array at the root of the high-pressure rotor impeller to collect internal temperature gradients, surface transient temperature fields, and micro-deformation characteristics, and generate a three-dimensional temperature and strain tensor set with spatial coordinates;

[0060] Step 2: Receive the three-dimensional temperature and strain tensor set, enhance the resolution of the curved surface area through a dynamic grid encryption strategy, combine metal phase transition parameters to reconstruct the temperature field and stress field in real time, and output a rasterized stress cloud map;

[0061] Step 3, analyze the stress accumulation rate and spatial distribution characteristics of the gridded stress cloud map, generate control signals including steam valve opening degree instructions and temperature adjustment amounts according to the thermal state risk level switching calculation model;

[0062] Step 4, receive the temperature adjustment amount in the control signal, identify the rotor vibration modal characteristics in real time, and dynamically generate the hydraulic damping force field and the throttle flow distribution strategy by matching the shaft system resonance speed prediction results;

[0063] Step 5, receive the steam valve opening degree instruction in the control signal, encode the steam pressure change trajectory, steam temperature gradient curve and drain valve action sequence into a chromosome structure during the shutdown stage, and optimize the cooling path through a multi-objective evolutionary algorithm.

[0064] Advantages of the present application

[0065] The present application is based on the embedded optical fiber matrix of axial and radial cross layout and the cooperative collection of curved surface attached infrared array, generates three-dimensional temperature and strain tensor set through laser space coordinate binding and atomic clock synchronization fusion, solves the problem of insufficient multi-source data fusion accuracy of high-pressure rotor curved surface structure; steam flow speed response type grid subdivision combined with phase change parameter dynamic loading realizes eight times resolution improvement in groove area, synchronously reconstructs heat transfer boundary through flow state mapping table, eliminates complex structure heat transfer hysteresis effect; digital twin driven crack prediction and measured position confidence closed loop correction mechanism dynamically calibrates model parameters, makes temperature field reconstruction error control within safety margin range, effectively suppresses the risk of thermal stress crack initiation; the cooling path optimized by evolutionary algorithm compresses the start-stop cycle of aging unit beyond the conventional system, cooperates with the phase reconstruction technology of vibration suppression module to reduce the resonance amplitude, and comprehensively improves the service life of the unit and the response ability of the power grid. BRIEF DESCRIPTION OF DRAWINGS

[0066] In order to more clearly illustrate the technical solutions of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. Obviously, for those skilled in the art, other drawings can also be obtained from the drawings without creative labor.

[0067] Figure 1 The flow chart of the steam turbine rapid start-stop control method provided by the embodiment of the present application. DETAILED DESCRIPTION

[0068] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described below in connection with specific embodiments of the present application and corresponding drawings. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application. The technical solutions provided by the embodiments of the present application will be described in detail below in connection with the drawings. In order to better understand the objects of the present application, the present application will be further described in detail.

[0069] In a first aspect, the present application provides a steam turbine rapid start-stop control system, comprising:

[0070] A high-precision sensing module is arranged at the root of the high-pressure rotor impeller to collect the internal temperature gradient, surface transient temperature field and micro-deformation characteristics of the metal, and generate a three-dimensional temperature and strain tensor set with spatial coordinates;

[0071] A thermal stress reconstruction module receives the three-dimensional temperature and strain tensor set, enhances the resolution of the curved surface area through a dynamic grid encryption strategy, reconstructs the temperature field and stress field in real time in combination with the metal phase change parameters, and outputs a rasterized stress cloud map;

[0072] An adaptive decision module analyzes the stress accumulation rate and spatial distribution characteristics of the rasterized stress cloud map, switches the calculation model according to the thermal state risk level, and generates control signals including the steam valve opening degree instruction and the temperature adjustment amount;

[0073] A vibration suppression execution module receives the temperature adjustment amount in the control signal, identifies the rotor vibration modal characteristics in real time, matches the shaft system resonance speed prediction result to dynamically generate a hydraulic damping force field and a throttle flow distribution strategy;

[0074] A cooling optimization execution module receives the steam valve opening degree instruction in the control signal, encodes the steam pressure change trajectory, steam temperature gradient curve and drain valve action sequence into a chromosome structure during the shutdown stage, and optimizes the cooling path through a multi-objective evolutionary algorithm;

[0075] The output of the high-precision sensing module drives the operation of the thermal stress reconstruction module;

[0076] The rasterized stress cloud map output by the thermal stress reconstruction module triggers the model scheduling of the adaptive decision module;

[0077] The adaptive decision module transmits the control signals in a channel: the temperature adjustment amount to the vibration suppression execution module, and the steam valve opening degree instruction to the cooling optimization execution module.

[0078] The high-precision sensing module is arranged with an intelligent sensing array that is axially and radially crossed positioned at the root groove area of the first to third stage impellers of the high-pressure rotor. The array deployment includes two complementary measurement units: a flexible optical fiber matrix embedded in the metal body collects the internal temperature gradient distribution, an infrared array installed in a curved surface profile synchronously captures the surface transient temperature field, and a network of piezoelectric micro-strain sensors captures the microscopic deformation fluctuations. The three sensing sources output heterogeneous raw data, which are fused by the time and space calibration engine to generate a temperature and strain tensor set with three-dimensional spatial coordinates. The tensor set represents the thermal coupling state of the rotor curved surface area, and the spatial coordinate accuracy is constrained by the laser positioning topology network, and the time synchronization is guaranteed by the atomic clock source.

[0079] The thermal stress reconstruction module receives the three-dimensional temperature and strain tensor set, and starts a multi-scale dynamic grid partitioning protocol. According to the real-time steam flow rate change rate, the grid density strategy is dynamically switched: in the high flow condition of flow rate > 15 m / s, recursive subdivision is implemented for the curvature mutation area of the blade root groove, and the grid size is shrunk to 1 / 8 of the basic unit; when the phase transition critical temperature interval (400-550℃) is detected, the austenite transformation characteristic parameter is called to update the local Young's modulus in the dynamic material property management subsystem. After grid optimization and material property matching, the steam flow state-heat transfer coefficient mapping table is loaded through the thermal and mechanical delay corrector to reconstruct the boundary conditions, and finally drive the finite element inversion model to output the gridded stress cloud map, with a spatial resolution of 1.5 mm 3 / grid in the stress concentration area.

[0080] When the adaptive decision-making module analyzes the gridded stress cloud map, it first extracts two key feature quantities: stress accumulation rate (MPa change per unit time) and spatial variation index. Based on the double feature quantities, a thermal state risk level evaluation matrix is constructed: when the axial temperature standard deviation is ≤5℃ and the variation index is <0.3, it is evaluated as low risk, and the simplified beam model calculation path is activated; when the groove area temperature rise rate is >10℃ / s or the variation index is >0.7, it is determined as high risk, and the high-frequency iterative shell element model is switched. The calculation model output end is connected to the spatial grid management subsystem, which discretizes the stress field into dynamic size cubic cells, automatically marks red warning for the plastic domain over-limit grid, and finally generates a double-channel control signal containing the steam valve opening degree instruction and the temperature regulation amount.

[0081] When the vibration suppression execution module receives the temperature adjustment amount signal, a multi-source signal fusion mechanism is started. A MEMS three-axis acceleration sensor cluster arranged circumferentially around the bearing seat collects real-time original vibration waveforms, and extracts bending or torsional modal weight coefficients through joint time-frequency domain analysis. The modal feature vector is input into the vibration suppression strategy generator along with the temperature adjustment amount, and when the temperature adjustment amount is greater than 30 DEG C, a thermal compensation algorithm is activated to reconstruct a phase distribution matrix. Two-way execution instructions are generated at the output end simultaneously: the hydraulic actuator array generates a reverse damping force field according to the reconstructed matrix, and the DEH governor controller reallocates the opening degree of the high-pressure governor group according to the flow-vibration sensitivity matrix, so as to realize thermal-mechanical dual vibration suppression.

[0082] After the cooling optimization execution module receives the steam valve opening degree instruction in the shutdown stage, a genetic evolution optimization process is started. The control instruction analysis unit extracts three characteristic parameters, i.e., the steam pressure drop trajectory, the steam temperature gradient curve and the time sequence of the drain valve action, from the opening degree instruction sequence. The gene coding system converts the parameter sequence into a chromosome gene segment, and the fitness evaluation system calculates the weighted function value of the stress suppression rate (peak stress drop percentage) and the shutdown time compression rate. The evolution execution system implements an elite reservation strategy, and reserves the top 20% of chromosomes in terms of stress safety factor in each generation. The control instruction step mutation is avoided through time-sharing gene crossing. The boundary supervisor forces to check the constraint condition that the main steam superheat degree is greater than or equal to 30 DEG C in the iteration process, and finally outputs the optimal cooling path that meets the rotor life state.

[0083] The whole system constructs a closed-loop feedback through a digital twin verification module. A data fusion gateway collects a gridized stress snapshot and a vibration spectrum to establish a four-dimensional space-time index, and a damage evolution prediction engine converts a local strain amplitude into a crack propagation rate curve according to a Manson-Coffin model. A parameter calibration microservice compares the endoscopic measured crack coordinates with the predicted position in real time, generates a confidence correction coefficient when the spatial offset is greater than 0.8 mm, and dynamically updates the prediction model parameter library. At the same time, the decision threshold is dynamically adjusted based on the residual life prediction value: the elastic domain warning threshold of the aging unit is 30%, so that the safety margin remains constant during the life decline period.

[0084] Specifically, the steam turbine rapid start-stop control system provided by the application comprises a high-precision sensing module, a vibration suppression execution module, a cooling optimization execution module and a digital twin verification module.

[0085] The embedded optical fiber matrix is arranged in an axial and radial cross layout mode in the blade root groove area.

[0086] The curved surface fitting infrared array covers the groove surface area and compensates for motion blur errors.

[0087] The collected data of the embedded optical fiber matrix and the curved surface fitting infrared array are input into a space-time calibration engine.

[0088] The spatio-temporal calibration engine binds sensor physical coordinates through a laser positioning system and synchronizes multi-source data timestamps, and outputs calibrated sensor data sets;

[0089] The calibrated sensor data sets drive an interpolation algorithm to generate the three-dimensional temperature field.

[0090] The embedded optical fiber matrix adopts a pre-set curvature flexible substrate packaging process to implement physical deployment. A profiled groove is pre-processed in the high-pressure rotor blade root groove area, an optical fiber Bragg grating array carried by a ceramic substrate is embedded, a continuous temperature measurement chain with a spacing of ≤35 mm is arranged in the axial direction, and redundant measurement points are arranged in the radial direction to form a double-channel verification topology. For high-speed rotating conditions, a bellows type metal sheath is assembled at the grating node to suppress centrifugal stress, and a gradient material transition layer is used to buffer thermal mechanical strain, so that the spatial resolution of internal temperature gradient measurement is improved to the core area of the curved structure.

[0091] The curved surface fitting infrared array is installed on the rotor surface through a tungsten alloy support. The curved mirror set covers all observation dead angles in the groove area, and the mirror curvature radius is accurately matched with the local geometric characteristics of the rotor. The miniature infrared sensor integrates a self-cleaning air curtain to isolate high-temperature steam interference, and a motion blur compensation algorithm is built-in to dynamically correct the thermal image frame integration time according to the real-time rotating speed of the rotor. At the same time, based on the metal emissivity-temperature characteristic curve library, the radiation characteristics of 304H stainless steel in the 400-600℃ interval are compensated online, so that the surface transient temperature field capture error is controlled within ±1.5℃.

[0092] The spatio-temporal calibration engine synchronously processes multi-source sensing signals. The laser positioning system arranges a multi-spectral emitter at the stationary end of the cylinder, and forms a dynamic coordinate grid on the rotor surface when rotating. The positioning chip built-in each sensor calculates the physical coordinates through triangular intersection. The main control cabinet is equipped with a rubidium atomic clock source, which broadcasts synchronization pulses to all sensing nodes through a fiber timing network, and writes millisecond-level timestamps in the header of the data frame. The edge computing node implements a clock offset compensation algorithm to eliminate signal transmission delays caused by rotating speed fluctuations, and finally outputs a sensor original data set with aligned spatio-temporal coordinates.

[0093] After receiving the calibrated data set, the interpolation algorithm module starts a hierarchical fusion process. First, the centrifugal force effect compensation coefficient is applied to the fiber temperature signal to correct the grating wavelength drift error; and the non-uniformity correction is performed on the infrared image to eliminate the mirror reflection interference. The temperature gradient field is reconstructed based on the spatial topological relationship of the sensors by using the Kriging spatial interpolation algorithm. The interpolation weight calculation introduces a local surface curvature factor, which automatically increases the weight coefficient of the adjacent measurement points in the curvature mutation area at the root of the groove, to generate a temperature and strain fusion tensor set with spatial three-dimensional coordinates.

[0094] Specifically, the steam turbine rapid start-stop control system provided by the application comprises a thermal stress reconstruction module.

[0095] a dynamic material property management subsystem storing Young's modulus parameters associated with austenite / ferrite phase transition temperatures;

[0096] an intelligent mesh subdivision engine receiving the three-dimensional temperature and strain tensor set, performing dynamic adjustment of mesh density according to a real-time steam flow rate change rate, recursively subdividing a mesh in a blade root groove region under a high flow rate working condition, and inserting a transition layer mesh in a curvature mutation area;

[0097] an output of the intelligent mesh subdivision engine being connected to a thermal delay correction system, the system loading the steam flow state mapping table to reconstruct a rotor surface heat exchange boundary condition and constraining an inversion calculation iteration convergence speed;

[0098] the updated boundary condition output by the thermal delay correction system being input into a grid inversion model of the dynamic material property management subsystem to generate a rasterized stress cloud map.

[0099] The dynamic material property management subsystem constructs a rotor metal phase transition behavior feature library. A Young's modulus decay curve corresponding to an austenite-ferrite transition critical temperature interval is stored, and a temperature-modulus gradient mapping table is established in the 400-550℃ phase transition active zone. When a local temperature entering the phase transition interval is detected during the reconstruction process, the corresponding material parameter curve of the region is automatically called, and the material response mode is switched through a state machine model. For CrMoV alloy rotor materials, the creep inflection point characteristics of the pearlite spheroidization stage above 550℃ are specially marked to realize accurate modeling of the nonlinear response of materials in the high-temperature zone.

[0100] After receiving the three-dimensional temperature and strain tensor set, the intelligent mesh subdivision engine starts a steam flow rate responsive mesh adjustment protocol. The main steam flow rate change gradient is monitored in real time, and when the flow rate increment exceeds 3m / s per second, the dynamic encryption strategy is activated: recursive binary subdivision is implemented in the blade root groove area, so that the local mesh size is reduced to one fourth of the basic unit; for the curvature radius mutation part, three-level transition layer meshes are automatically inserted to eliminate the calculation singular point in the stress concentration area. The grid topology optimizer synchronously scans the model geometric features, and the continuous and gradual mesh distribution is forced to be maintained in the fillet area with a radius mutation of more than 20%.

[0101] The thermal delay correction system is connected to the output end of the mesh subdivision engine, and loads a steam flow state-heat exchange coefficient mapping relationship library. The flow state is identified according to the multi-section flow probe data arranged in the steam pipeline, and when the Reynolds number exceeds the critical value, the turbulent heat exchange model is switched to update the boundary condition. The calculation stability controller monitors the variable oscillation amplitude in the inversion iteration process in real time, and if the stress difference of two consecutive iterations exceeds 5%, the Lyapunov index adaptive adjustment is triggered, the iteration step is shrunk to 60% of the previous value, and the reduced model is reconstructed to converge the path.

[0102] The grid inversion model fuses multi-source updated parameters to perform thermal stress reconstruction. The phase change region modulus parameters of the dynamic material attribute library and the new boundary conditions of the thermal force correction system are synchronously received on the input side, and weighted finite element solving is implemented under a parallel computing architecture. Local thermal force balance verification is implemented at each calculation step, local grid reconstruction is triggered when the temperature gradient difference of adjacent elements in the groove region exceeds a set threshold, and finally a grid stress cloud map with a spatial resolution of 1mm 3 is output.

[0103] Specifically, the steam turbine rapid start-stop control system comprises a self-adaptive decision module.

[0104] A dynamic model routing gateway receives the grid stress cloud map, extracts stress accumulation rate and spatial distribution characteristics, and performs calculation model switching according to the thermal state risk level: a simplified stress calculation model is called when the axial temperature standard deviation meets a first threshold, and a high-frequency iteration model is activated when the groove region temperature rise rate exceeds a second threshold.

[0105] The stress evaluation results output by the dynamic model routing gateway are input into a spatial grid management subsystem, which discretizes the stress field into variable-size cubic elements, automatically encrypts the grid distribution in the high-stress area and marks the elastic domain or plastic domain double-threshold state, and generates a risk assessment signal feedback to the dynamic model routing gateway.

[0106] The risk assessment signal of the spatial grid management subsystem drives the generation of control signals.

[0107] After the dynamic model routing gateway receives the grid stress cloud map, it first performs a key feature parameter extraction algorithm. A spatial gradient operator is used to scan the stress field distribution, and the stress accumulation rate per unit time of each monitoring point is calculated. At the same time, an axial temperature distribution histogram is constructed to quantify the uniformity of the temperature field. The temperature rise gradient of the groove region is tracked, and the peak point of the temperature rise rate is marked. The feature extraction results are input into a thermal state risk assessment matrix: when the axial temperature standard deviation is ≤4℃ and the stress rate is <5MPa / min, the beam element simplified model is activated for low risk level; when the groove temperature rise rate is >8℃ / s or the stress rate is >15MPa / min, the shell element high-frequency iteration model is switched for high risk level, and the model calculation frequency is positively correlated with the risk level.

[0108] After the spatial grid management subsystem receives the stress evaluation results, it starts an intelligent discretization process. The rotor three-dimensional model is initially divided into 5mm basic cubic elements, and the spatial coordinates of the high stress area are dynamically scanned. The recursive subdivision mechanism is triggered for the area with a stress value exceeding 80% of the elastic limit, and the grid density is improved to 1mm 3Each grid implements a double-threshold marking protocol: the elastic domain threshold is set to 70% of the material yield strength, and the plastic domain threshold is set to 90%, which are marked with yellow and red colors respectively to identify the risk level. The generated grid risk distribution map is fed back to the routing gateway model calculation node in real time, triggering the model switching mechanism to update.

[0109] The control signal generation engine dynamically constructs the output parameters according to the risk assessment signal. The risk distribution heat map pushed by the space grid management subsystem is analyzed to determine the spatial clustering state of the yellow and red grids. For the condition that the total volume of continuous red grid area is greater than 3%, a step load reduction instruction for the steam valve opening degree is generated; when the high-risk grid and the temperature abnormal area are spatially overlapped, the temperature regulation amount output proportion is increased synchronously. In the final generated double-channel control signal, the steam valve opening degree instruction maps the risk grid density change curve, and the temperature regulation amount is associated with the weighted function of the axial temperature standard deviation and the groove temperature rise rate.

[0110] Specifically, the steam turbine rapid start-stop control system provided by the application comprises:

[0111] A vibration sensing unit is arranged in a spatial spiral layout mode to arrange a MEMS sensor cluster, collect rotor three-axis acceleration characteristics, and generate original vibration signals;

[0112] A resonance prediction engine receives the original vibration signals for joint time-frequency domain analysis, compares historical vibration mode atlas libraries to identify bending or torsional modal characteristic vectors;

[0113] The output of the resonance prediction engine is connected to a vibration suppression strategy execution unit, which simultaneously receives the temperature regulation amount in the control signal, dynamically allocates the phase angle of the hydraulic actuator according to the modal characteristic vector and the temperature regulation parameter, and adjusts the opening ratio of the DEH system high-pressure control valve group and the steam supplement valve.

[0114] The vibration sensing unit implements a spatial spiral control strategy in the key areas of the bearing seat and the shaft neck. MEMS three-axis acceleration sensor clusters are installed at an angle of 45° along the rotor axis, and a three-dimensional vibration capture network is formed by the spiral layout with a phase difference of 90°. The ceramic packaging shell is resistant to oil and gas erosion environment, and the signal transmission uses twisted shielded cable to resist electromagnetic interference. After the sensor output original vibration waveform is filtered by an anti-aliasing filter, a digital signal including amplitude / frequency / phase components is generated, and the time accuracy is aligned to the level of hundreds of nanoseconds through an optical fiber timing system.

[0115] The resonance prediction engine starts the time-frequency domain joint analysis process after receiving the digital signal. The wavelet packet transform algorithm is applied to decompose the frequency band energy distribution of the vibration signal, and the characteristic harmonic components in the 0.5-3 times of the power frequency range are extracted. The shaft system modal fingerprint stored in the historical start-stop database is compared, and when the first-order bending modal energy accounts for more than 40% of the total energy or the phase mutation of the second-order torsional modal is more than 15°, it is determined that the resonance risk state is output. The modal characteristic vector includes three key parameters: the main vibration direction angle, the modal weight coefficient and the frequency band energy concentration.

[0116] The vibration suppression strategy execution unit synchronously receives the modal characteristic vector and the temperature regulation signal. When the temperature regulation amount exceeds the 30℃ threshold, the thermal deformation compensation algorithm is activated, and the hydraulic actuator phase distribution matrix is reconstructed: for the bending mode, the radial damping force phase component is enhanced, and for the torsional mode, the tangential damping force proportion is increased. According to the mode-governor sensitivity matrix, the steam flow is dynamically distributed: when the first-order bending mode is dominant, the high-pressure governor group opening is reduced and the steam valve flow proportion is increased; when the second-order torsional mode is significant, the medium-pressure governor throttling proportion is increased. The output end synchronously sends the pulsating pressure command to the hydraulic actuator array and transmits the governor cooperative action sequence to the DEH system.

[0117] Specifically, the steam turbine rapid start-stop control system described in the present application comprises:

[0118] A control instruction analysis unit receives the steam valve opening instruction, extracts the steam pressure change trajectory, steam temperature gradient curve and drain valve action sequence;

[0119] The output of the control instruction analysis unit is connected to a gene coding system, which encapsulates the extracted steam pressure change trajectory, steam temperature gradient curve and drain valve action sequence into a chromosome structure;

[0120] The chromosome structure of the gene coding system is input into a fitness evaluation system, which executes a multi-objective evolutionary algorithm calculation process: calculates the weighted evaluation value of stress peak suppression rate, superheat degree maintenance index and shutdown time compression rate;

[0121] The output of the fitness evaluation system drives the evolutionary execution system to run, which adopts a time-periodic gene crossover strategy to recombine the control chain sequence, and in the iteration process, the main steam superheat degree boundary constraint is forcibly checked, and the optimized cooling path instruction is output.

[0122] The control instruction analysis unit starts the parameter extraction process after receiving the steam valve opening degree instruction. The steam pressure drop rate change trajectory is parsed from the opening degree instruction time sequence, and the pressure inflection point coordinates are identified; the steam temperature gradient change curve of the temperature control channel is extracted synchronously, and the critical temperature drop interval is marked; the drain valve action sequence is parsed into the on-off state time sequence table through the instruction operation code. The parameter extraction algorithm implements instruction semantic analysis, maps the opening degree value to the physical quantity change trajectory, and establishes the corresponding relationship matrix of time-pressure / temperature.

[0123] The genetic coding system receives the parameter trajectory data and performs chromosome packaging operation. The steam pressure change trajectory is decomposed into a time-pressure value key-value pair sequence as the first gene segment of the chromosome; the steam temperature gradient curve is converted into a set of temperature inflection point coordinates to form the second gene segment; and the drain valve action sequence is coded into a binary on-off state matrix to form the third gene segment. The three-segment gene structure adopts a binary and floating-point number hybrid coding mode, and 100 groups of parameter combinations are generated by orthogonal test design to ensure gene diversity.

[0124] The fitness evaluation system performs a multi-objective evolutionary calculation process. The control parameters are decoded from the chromosome structure, and three optimization objectives are calculated: stress peak suppression rate, overheat maintenance index, and shutdown time compression rate. The multi-objective weighted function dynamically allocates weight coefficients according to the rotor life state: the weight of the time compression rate is increased for new units, and the weight of the stress suppression rate is increased for aging units. The fitness evaluation value is used as the evolution selection criterion.

[0125] The evolution execution system adopts an elite reservation strategy to implement iterative optimization. The population of each generation is sorted by stress safety factor, and the top 20% elite chromosomes are reserved for the next generation. The time-segmented gene crossover algorithm recombines the steam pressure / steam temperature control chain in different intervals to maintain the integrity of the control logic in each interval while avoiding step mutations. The boundary constraint supervisor checks the condition that the main steam overheat is greater than or equal to 30°C after each iteration, and the out-of-bound individuals trigger the re-encoding mechanism. Finally, the optimized cooling path instruction sequence that meets multiple constraints is output.

[0126] Specifically, the steam turbine rapid start-stop control system described in the present application, during the operation of the cooling optimization execution module,

[0127] The fitness evaluation system obtains the rotor life state parameters in real time, and executes a dynamic weight allocation strategy according to the rotor life state parameters:

[0128] Under the working condition of a new unit, a high weight is allocated to the shutdown time compression rate;

[0129] Under the working condition of an aging unit, a high weight is allocated to the stress peak suppression rate;

[0130] The output of the weight allocation strategy is fed back to the evolutionary execution system in real time, driving the adjustment of the gene crossover strategy in the iteration process.

[0131] The rotor life state parameters are obtained through the historical overhaul database and real-time monitoring system. The cumulative operating hours are used as the basic life index, combined with the start-stop cycle number to calculate the fatigue damage factor, and the number of microscopic cracks found by endoscopic detection to construct the life state evaluation vector. When the operating hours are < 50,000 hours and the damage factor is < 0.3, it is classified as a new unit operating condition; when the cumulative operating hours are > 200,000 hours or the number of cracks is ≥ 5 per rotor, it is determined as an aging unit operating condition, and a state recognition code is generated and transmitted to the weight allocator.

[0132] The weight allocator receives the life state code and executes a dynamic mapping strategy. In the new unit operating condition, the weight coefficient of the shutdown time compression rate is increased to 70% of the total weight, the stress peak suppression rate weight is reduced to 20%, and the overheat maintenance index remains at 10% of the basic weight. In the aging unit operating condition, the weight proportion is configured in reverse: the stress peak suppression rate weight accounts for 65%, and the shutdown time compression rate weight is compressed to 25%. The weight ratio calculator generates a ternary coefficient vector (W1, W2, W3) in real time according to the operating condition;

[0133] The embedded fitness evaluation formula is: Fitness = W1·S r +W2·T t +W3·O s (S r : stress suppression rate / T t : time compression rate / O s : overheat index).

[0134] The gene crossover strategy adjustment module receives the weight coefficient vector and drives the evolutionary mechanism optimization. When the high weight is allocated to the shutdown time compression rate (new unit mode), the accelerated evolution protocol is activated: the probability threshold of gene crossover operation is expanded to 35%, and the control chain timing reorganization density is increased. When the stress peak suppression rate occupies the dominant weight (aging unit mode), switch to the conservative evolution strategy: increase the elite retention ratio to 30% of the population size, and add a material crack resistance screening operator at the gene crossover link to preferentially retain chromosome sequences with high stress safety factor.

[0135] The feedback execution controller implements the closed-loop linkage of weight strategy and genetic operation. The current weight coefficient vector is loaded before each generation of population evolution iteration, and the crossover operator parameter library is reset synchronously when the weight configuration is switched. In the iteration process, the optimization target change rate of adjacent generations is compared in real time. If the improvement amplitude of time compression rate or stress suppression rate is less than 5% for two consecutive generations, the adaptive reset of crossover strategy parameters is triggered, and the crossover probability and mutation intensity coefficient are updated by gradient descent method.

[0136] Specifically, the steam turbine fast start-stop control system also includes a closed-loop verified digital twin module:

[0137] A data fusion gateway receives the gridded stress cloud map of the thermal stress reconstruction module and the vibration frequency spectrum data of the vibration suppression execution module.

[0138] A holographic data warehouse system connected to the output of the data fusion gateway stores gridded stress snapshots and vibration frequency spectra according to a four-dimensional space-time index.

[0139] A damage evolution prediction engine loads historical stress data from the holographic data warehouse system and calculates a crack propagation curve based on local strain amplitude.

[0140] A parameter calibration microservice receives the crack propagation curve and residual life prediction value from the damage evolution prediction engine and feeds the residual life percentage of the rotor back to the threshold management system of the adaptive decision module.

[0141] The data fusion gateway establishes a multi-source data access channel. When receiving the gridded stress cloud map of the thermal stress reconstruction module, it implements spatial coordinate alignment, binding the three-dimensional grid data in the stress cloud map with the physical coordinates of the rotor. It also synchronously accesses the frequency domain characteristic signals of the vibration suppression execution module and extracts the main frequency band energy distribution value of the vibration frequency spectrum through a time scale analysis engine. The gateway has a built-in data format standardization protocol that encapsulates the spatial distribution matrix of the stress field and the frequency energy vector of the vibration into a unified space-time frame structure, with a time stamp accuracy matching the hundred-millisecond control cycle.

[0142] The holographic data warehouse system implements a four-dimensional indexing architecture. The space-time coordinate system constructs a four-tuple index key with start-stop cycle number, run-time stamp, rotor radial angle, and axial position. Gridded stress snapshots are stored in blocks according to spatial coordinates, and vibration frequency spectrum data are archived in frequency bands. The historical data compression algorithm automatically identifies low-frequency change areas for lossy compression and high-frequency mutation areas for lossless data format. The storage engine supports fast retrieval by space-time range, such as extracting the stress distribution pattern in a specific speed range during the last 10 start-stop processes.

[0143] The damage evolution prediction engine loads historical strain data to reconstruct the crack evolution path. Based on the Manson-Coffin fatigue damage model, it converts the local strain amplitude sequence extracted from the warehouse system into equivalent plastic strain accumulation. For high-pressure rotor blade root groove areas, the Forman crack propagation rate equation is preferentially applied to calculate the crack length increment under a given stress intensity factor. The prediction engine outputs two characteristic curves: a crack depth curve changing with the number of start-stop cycles and a residual life curve decaying with crack propagation.

[0144] The parameter calibration microservice constructs a bidirectional feedback channel. After receiving the crack propagation curve data, the predicted crack initiation position coordinates and the remaining life percentage value are extracted. The remaining life parameter is input into the threshold management system of the adaptive decision module. When the life value is lower than 50%, the safety margin compression protocol is started: the elastic domain warning threshold is lowered from 70% of the material yield strength to 60%. The crack position coordinates and the endoscopic measurement results are subjected to spatial deviation analysis. When the coordinate offset exceeds the allowed value, a confidence correction coefficient is generated and fed back to the damage prediction engine to update the critical strain threshold parameter.

[0145] Specifically, the steam turbine rapid start-stop control system provided by the application establishes a dynamic feedback channel through the parameter calibration microservice:

[0146] After receiving the crack propagation curve and the remaining life prediction value of the damage evolution prediction engine,

[0147] When the remaining life prediction value is lower than the set threshold,

[0148] A threshold contraction instruction is generated and transmitted to the spatial grid management subsystem of the adaptive decision module,

[0149] The elastic domain warning threshold is adjusted downward to the preset safety margin;

[0150] An endoscopic measurement data acquisition unit acquires a rotor surface crack distribution image and inputs it into the parameter calibration microservice;

[0151] The parameter calibration microservice performs crack position deviation analysis:

[0152] By comparing the coordinates of the crack distribution image and the predicted crack position, the spatial position offset is calculated, and a confidence correction coefficient is generated according to the offset;

[0153] The confidence correction coefficient is input into the damage evolution prediction engine to update the calculation parameters of the crack propagation curve.

[0154] When the parameter calibration microservice initializes the dynamic feedback channel, it first loads the key data set output by the damage evolution prediction engine. The spatial coordinates of the predicted crack initiation position in the crack propagation curve are analyzed, and the remaining life prediction percentage value is read synchronously. When the life prediction value is lower than the preset 50% threshold, the safety margin compression protocol is activated: a threshold contraction instruction is generated and transmitted to the spatial grid management subsystem of the adaptive decision module through a standard interface, triggering the elastic domain warning threshold to be lowered from the 70% reference value of the material yield strength to the set safety margin interval, and the lower limit value of the interval is determined by the fatigue limit characteristics of the rotor material.

[0155] The endoscope measurement data acquisition unit performs surface crack scanning at the unit shutdown window. The high-definition surface image of the high-pressure rotor blade root groove area is obtained by using an industrial endoscope imaging system, the actual crack distribution pattern is identified and the spatial coordinates are marked by a feature extraction algorithm. The measured coordinate data set is converted to a coordinate system matching the prediction model by a topological mapping algorithm, and a standardized crack distribution map input parameter is generated to calibrate the microservice.

[0156] The crack position deviation analysis module performs a quantitative verification process. After receiving the endoscope measured crack map and the predicted position coordinates, a three-coordinate space transformation model is used to implement point set registration alignment. The average Euclidean distance between the predicted coordinate points and the actual crack points is calculated, and when the spatial offset exceeds 0.8mm, it is determined that the deviation is out of limit. The offset quantitative value is input into the confidence conversion function to generate a dynamic correction coefficient in the range of 0.8-1.2, which is negatively correlated with the offset.

[0157] The damage prediction engine parameter updating mechanism optimizes the model in real time according to the confidence coefficient. The correction coefficient is injected into the crack propagation calculation module through the application programming interface, and the critical strain amplitude parameter in the Forman equation is adjusted in proportion. When the correction coefficient is less than 1.0, the crack propagation rate calculation model automatically increases the local stress sensitivity weight; at the same time, the coordinate system conversion parameter library is iteratively updated according to the historical deviation record, and the cumulative error of space mapping is reduced.

[0158] In a second aspect, referring to Figure 1 , the application provides a steam turbine rapid start-stop control method, which is applied to the steam turbine rapid start-stop control method, comprising:

[0159] Step 1, deploying a sensor array at the root of the high-pressure rotor impeller, collecting the internal temperature gradient, surface transient temperature field and micro-deformation characteristics of the metal, and generating a three-dimensional temperature and strain tensor set with spatial coordinates;

[0160] Step 2, receiving the three-dimensional temperature and strain tensor set, enhancing the resolution of the curved surface area by a dynamic grid encryption strategy, combining the metal phase change parameters to reconstruct the temperature field and stress field in real time, and outputting a gridded stress cloud map;

[0161] Step 3, analyzing the stress accumulation rate and spatial distribution characteristics of the gridded stress cloud map, generating control signals including steam valve opening degree instructions and temperature adjustment amounts according to the thermal state risk level switching calculation model;

[0162] Step 4, receiving the temperature adjustment amount in the control signal, identifying the rotor vibration modal characteristics in real time, matching the shaft system resonance speed prediction result to dynamically generate a hydraulic damping force field and a throttle flow distribution strategy;

[0163] Step 5, receiving the steam valve opening instruction in the control signal, encoding the steam pressure change trajectory, steam temperature gradient curve and steam trap action sequence into a chromosome structure in the shutdown stage, and optimizing the cooling path through a multi-objective evolutionary algorithm.

[0164] The present application solves the problem of temperature field distortion through multi-modal collaborative sensing and dynamic adaptive reconstruction strategy. In the root groove area of the high-pressure rotor impeller, an embedded optical fiber matrix and a curved surface attached infrared array with axial / radial cross layout are arranged. The former captures the internal temperature gradient distribution of the metal, and the latter synchronously acquires the surface transient temperature field. The space-time calibration engine fuses the temperature and strain tensor set with three-dimensional space coordinates through laser coordinate binding and atomic clock synchronization, and establishes a high-precision space-time reference.

[0165] The thermal stress reconstruction module adopts a working condition response type grid division protocol to eliminate calculation distortion. Real-time monitoring of steam flow rate change rate, when the flow rate increase amplitude exceeds the threshold, the grid recursive subdivision mechanism is automatically triggered for the groove curvature mutation area, and the local grid resolution is improved by eight times. At the same time, the rotor material phase change parameter library is dynamically loaded, and the local Young's modulus is dynamically updated in the austenite transformation temperature range to compensate for the material response error caused by thermal hysteresis effect. The thermal force delay correction system reconstructs the heat transfer boundary conditions according to the steam flow state mapping table, and constrains the inversion iterative convergence speed.

[0166] The closed-loop verification mechanism continuously optimizes the reconstruction accuracy. The digital twin module stores the historical stress cloud and vibration spectrum through four-dimensional space-time indexing, and the damage evolution prediction engine updates the crack propagation model based on the local strain amplitude. The spatial offset between the endoscopic measured data and the predicted position coordinates drives the confidence correction coefficient generation, which is dynamically fed back to the reconstruction algorithm parameter library, realizing the adaptive calibration of the temperature field reconstruction model.

Claims

1. A rapid start-stop control system for a steam turbine, characterized in that, include: The high-precision sensing module deploys a sensor array at the root of the high-pressure rotor impeller to collect the internal temperature gradient of the metal, the transient temperature field of the surface, and the micro-deformation characteristics, and generate a three-dimensional temperature and strain tensor set with spatial coordinates. The thermal stress reconstruction module receives the three-dimensional temperature and strain tensor set, enhances the resolution of the curved surface region through a dynamic mesh densification strategy, and reconstructs the temperature field and stress field in real time by combining metal phase transformation parameters, and outputs a rasterized stress cloud map. The adaptive decision-making module analyzes the stress accumulation rate and spatial distribution characteristics of the gridded stress cloud map, and generates control signals including steam valve opening commands and temperature adjustment amounts based on the thermal state risk level switching calculation model. The vibration suppression execution module receives the temperature adjustment amount in the control signal, identifies the rotor vibration mode characteristics in real time, and dynamically generates a hydraulic damping force field and valve flow distribution strategy by matching the shaft resonance speed prediction results. The cooling optimization execution module receives the steam valve opening command in the control signal, encodes the steam pressure change trajectory, steam temperature gradient curve and condensate valve action sequence into a chromosome structure during the shutdown phase, and optimizes the cooling path through a multi-objective evolutionary algorithm. Wherein: the output of the high-precision sensing module drives the operation of the thermal stress reconstruction module; The rasterized stress cloud map output by the thermal stress reconstruction module triggers model scheduling of the adaptive decision module; The adaptive decision-making module transmits control signals through multiple channels: temperature adjustment to the vibration suppression execution module, and steam valve opening command to the cooling optimization execution module.

2. The turbine rapid start-stop control system according to claim 1, characterized in that, The high-precision sensing module includes: An embedded fiber optic matrix is ​​arranged in a cross-axial and radial layout in the blade root groove area; A curved surface-fitting infrared array covers the surface area of ​​the groove and compensates for motion blur errors; The data collected by the embedded fiber optic matrix and the curved surface-fitted infrared array are input to the spatiotemporal calibration engine; The spatiotemporal calibration engine binds the physical coordinates of the sensors through a laser positioning system and synchronizes the timestamps of multi-source data to output the calibrated sensor dataset. The calibrated sensor dataset drives the interpolation algorithm to generate the three-dimensional temperature field.

3. The turbine rapid start-stop control system according to claim 2, characterized in that, The thermal stress reconstruction module includes: The dynamic material property management subsystem stores Young's modulus parameters related to the austenite / ferrite phase transformation temperature; The intelligent mesh generation engine receives the three-dimensional temperature and strain tensor set, performs dynamic adjustment of mesh density according to the real-time steam velocity change rate, recursively subdivides the mesh in the blade root groove region under high flow rate conditions, and inserts transition layer meshes in the curvature abrupt change region. The output of the intelligent mesh generation engine is connected to the thermal delay correction system, which loads the steam flow regime mapping table to reconstruct the heat transfer boundary conditions on the rotor surface and constrains the convergence speed of the inversion calculation iteration. The updated boundary conditions output by the thermal delay correction system are combined with the parameter input mesh inversion model of the dynamic material property management subsystem to generate a rasterized stress cloud map.

4. The turbine rapid start-stop control system according to claim 3, characterized in that, The adaptive decision-making module includes: The dynamic model routing gateway receives the gridded stress cloud map, extracts the stress accumulation rate and spatial distribution characteristics, and performs calculation model switching according to the thermal state risk level: when the axial temperature standard deviation meets the first threshold, the simplified stress calculation model is called; when the temperature rise rate of the groove area exceeds the second threshold, the high-frequency iterative model is activated. The stress assessment results output by the dynamic model routing gateway are input to the spatial grid management subsystem. The system discretizes the stress field into variable-size cubic units, automatically densifies the grid distribution in high-stress areas and marks the double threshold states of the elastic or plastic domains, and generates a risk assessment signal that is fed back to the dynamic model routing gateway. The risk assessment signal drives the generation of control signals in the spatial grid management subsystem.

5. The turbine rapid start-stop control system according to claim 4, characterized in that, The vibration suppression execution module includes: The vibration sensing unit uses a cluster of MEMS sensors arranged in a spatial spiral layout to collect the triaxial acceleration characteristics of the rotor and generate raw vibration signals. The resonance prediction engine receives the original vibration signal and performs joint time-frequency domain analysis, comparing it with the historical mode shape library to identify the bending / torsional mode feature vectors. The output of the resonance prediction engine is connected to the vibration suppression strategy execution unit. The vibration suppression strategy execution unit simultaneously receives the temperature adjustment amount in the control signal, dynamically allocates the phase angle of the hydraulic actuator based on the temperature adjustment parameters fused by the modal feature vector, and adjusts the opening ratio of the high-pressure regulating valve group and the steam replenishment valve of the DEH system.

6. The turbine rapid start-stop control system according to claim 5, characterized in that, The cooling optimization execution module includes: The control command parsing unit receives the steam valve opening command and extracts the steam pressure change trajectory, steam temperature gradient curve, and steam trap action sequence. The output of the control command parsing unit is connected to the gene coding system, which encapsulates the extracted steam pressure change trajectory, steam temperature gradient curve and steam trap action sequence into a chromosome structure. The chromosome structure of the gene coding system is input into the fitness evaluation system, which executes a multi-objective evolutionary algorithm calculation process: calculating the weighted evaluation value of peak stress inhibition rate, overheat maintenance index and downtime compression rate; The output of the fitness evaluation system drives the evolutionary execution system. This system uses a time-segmented gene crossover strategy to recombine the control chain sequence, forcibly verifies the main steam superheat boundary constraints during the iteration process, and outputs optimized cooling path instructions.

7. The turbine rapid start-stop control system according to claim 6, characterized in that, During the operation of the cooling optimization execution module, The fitness evaluation system acquires rotor life state parameters in real time and executes a dynamic weight allocation strategy based on these parameters. Under the new unit operating conditions, a high weight is assigned to the downtime compression rate; Under aging unit operating conditions, a high weight is allocated to the stress peak suppression rate; The output of the weight allocation strategy is fed back to the evolutionary execution system in real time, driving the adjustment of the gene crossover strategy during the iteration process.

8. The turbine rapid start-stop control system according to claim 7, characterized in that, It also includes a digital twin module for closed-loop verification: The data fusion gateway receives the rasterized stress cloud map from the thermal stress reconstruction module and the vibration spectrum data from the vibration suppression execution module. The holographic data storage system, connected to the output of the data fusion gateway, stores rasterized stress snapshots and vibration spectra according to a four-dimensional spatiotemporal index. The damage evolution prediction engine loads historical stress data from the holographic data storage system and calculates crack propagation curves based on local strain amplitudes. The parameter calibration microservice receives the crack propagation curve and remaining life prediction value from the damage evolution prediction engine, and feeds back the rotor remaining life percentage to the threshold management system of the adaptive decision module.

9. The turbine rapid start-stop control system according to claim 8, characterized in that, The parameter calibration microservice establishes a dynamic feedback channel: After receiving the crack propagation curve and remaining lifetime prediction value from the damage evolution prediction engine, when the remaining lifetime prediction value is lower than a set threshold, A threshold shrinkage command is generated and transmitted to the spatial grid management subsystem of the adaptive decision module to adjust the elastic domain early warning threshold downward to a preset safety margin; The endoscopic measurement data acquisition unit acquires images of the crack distribution on the rotor surface and inputs them into the parameter calibration microservice. The parameter calibration microservice performs crack location deviation analysis: By comparing the coordinates of the crack distribution image with the predicted crack location, the spatial position offset is calculated, and a confidence correction coefficient is generated based on the offset. The confidence correction coefficient is input into the damage evolution prediction engine to update the calculation parameters of the crack propagation curve.

10. A method for rapid start-stop control of a steam turbine, applied to the rapid start-stop control method for a steam turbine as described in claims 1 to 9, characterized in that, include: Step 1: Deploy a sensor array at the root of the high-pressure rotor impeller to collect the internal temperature gradient of the metal, the transient temperature field of the surface, and the micro-deformation characteristics, and generate a three-dimensional temperature and strain tensor set with spatial coordinates. Step 2: Receive the three-dimensional temperature and strain tensor set, enhance the resolution of the curved surface region through a dynamic mesh densification strategy, reconstruct the temperature field and stress field in real time by combining metal phase transformation parameters, and output a rasterized stress cloud map. Step 3: Analyze the stress accumulation rate and spatial distribution characteristics of the gridded stress cloud map, and generate control signals including steam valve opening commands and temperature adjustment amounts based on the thermal state risk level switching calculation model. Step 4: Receive the temperature adjustment amount in the control signal, identify the rotor vibration mode characteristics in real time, and dynamically generate the hydraulic damping force field and valve flow distribution strategy by matching the shaft resonance speed prediction results. Step 5: Receive the steam valve opening command in the control signal, and encode the steam pressure change trajectory, steam temperature gradient curve and steam trap action sequence into a chromosome structure during the shutdown phase. Optimize the cooling path through a multi-objective evolutionary algorithm.

Images