Digital-based power plant state data processing method and system, and medium

Abstract

The application discloses a kind of based on digitization's thermal power plant state data processing method, system and medium, the method includes: obtaining whole-process distributed monitoring flow data and constructing original digital state space;Nonlinear coupling coefficient between parameters is solved using time-delay mutual information algorithm and adaptively pruned dimensionality, and dynamic nonlinear state topology graph is generated;According to thermodynamic boundary, the topology graph is decoupled and segmented into state cell matrix;The microcosmic parameter of each state cell unit is independently extracted and input into physical information neural network, and the continuity of space-time autocorrelation health evolution trajectory and damage critical threshold matrix is solved using the conservation law of multiple physical fields;Future load instruction is input into space-time collaborative digital twin evolution field for multi-scenario forward-looking deduction;Pareto optimization is carried out with energy efficiency optimization and component life maximization as collaborative control target.The application realizes the proactive closed-loop collaborative regulation of thermal power plant energy efficiency and life.

Description

Technical Field

[0001] This invention relates to the field of thermal power plant status data processing technology, and more specifically, to a digital thermal power plant status data processing method, system, and medium. Background Technology

[0002] With the deepening of the digital transformation of the power system and the construction of smart power plants, distributed control systems and plant-level information monitoring systems have deployed extremely high-density sensor networks throughout the plant. Thermal power plants, as typical large-scale dynamic thermodynamic industrial systems with large time delays, strong coupling, and high nonlinearity, continuously generate massive amounts of multi-dimensional, high-frequency industrial flow data daily. How to efficiently and accurately process this industrial big data and transform it into control decisions that can directly guide the optimization of plant-wide energy efficiency and the lifespan management of core components is currently a research hotspot in the field of intelligent control.

[0003] However, existing technical solutions have limitations: their multidimensional state flow data generally suffers from "nonlinear distortion" during dimensionality reduction. Existing DCS data integration and compression methods, in order to reduce the computational load on the central processing unit, typically employ a relatively coarse static linear tuning scheme. This involves manually selecting a primary parameter and forcing other auxiliary parameters to be spatially linearly correlated with that primary parameter. This forced linear alignment smooths out key nonlinear operating condition characteristics such as sudden changes in local thermal stress and transient turbulence disturbances, easily leading to severe distortion in subsequent digital monitoring and condition diagnosis during variable operating conditions and deep peak-shaving phases. Furthermore, there is a deep "reactive" disconnect between condition assessment and operational control decisions. Existing solutions typically treat core equipment as a macroscopic "black box," calculating macroscopic efficiency only based on total coal consumption and final work done over a long period, representing a typical passive maintenance approach. It not only fails to perceive the continuous thermal stress fatigue cycle and local entropy accumulation at the microscopic spatiotemporal level of high-risk components inside core equipment, but also fails to guide the distributed control system to implement proactive control parameter fine-tuning during operation. It is difficult to use algorithm scheduling to proactively delay the physical degradation of core hardware, thus significantly affecting the economic efficiency and safety of unit operation. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a digital-based method, system and medium for processing thermal power plant status data to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A digital-based method for processing status data of thermal power plants includes the following steps: S1 acquires distributed monitoring stream data of the entire process of thermal power plants in real time and constructs an original digital state space containing multiple parameter nodes. S2, calculate the dynamic nonlinear coupling coefficient between any two parameter nodes in the original digital state space based on the time-delay mutual information algorithm, and prune and reduce the dimensions of edges that do not meet the preset association threshold according to the dynamic nonlinear coupling coefficient, so as to generate a dynamic nonlinear state topology graph that characterizes the real-time operating conditions of the thermal power plant. S3, based on the topological boundaries and thermodynamic functional division of the physical equipment of the thermal power plant, the dynamic nonlinear state topology diagram is decomposed into multiple collaboratively related digital state cell units to form a state cell matrix. S4. For each of the digital state cell units, the local microscopic thermodynamic parameters, vibration spectrum parameters and material stress parameters are extracted independently. Based on the physical information neural network, the local entropy growth rate and the cumulative damage degree of microscopic thermal stress fatigue of each of the digital state cell units are calculated, and a continuous spatiotemporal autocorrelation health evolution trajectory is output to establish the damage critical threshold matrix of each of the digital state cell units. S5, receive the target load instruction requirements to be executed from the power grid, and input the target load instruction requirements to be executed as boundary conditions into the spatiotemporal collaborative digital twin evolution field composed of the state cell matrix to perform multi-scenario forward-looking evolution simulation, and simulate the nonlinear state transition process of the state cell matrix in the future time window. S6, during the multi-scenario forward-looking evolution simulation, with the optimal fuel economy of the entire plant and the maximization of the health and lifespan of each of the digital state cell units as the collaborative control objectives, the optimal load allocation matrix and dynamic control fine-tuning parameter flow of each subsystem of the entire plant are dynamically calculated without exceeding the damage critical threshold of any of the digital state cell units, and the optimal load allocation matrix and the dynamic control fine-tuning parameter flow are sent to the distributed control system for execution.

[0006] In some embodiments, S2 specifically includes: Define the discrete time series corresponding to any two parameter nodes in the original digital state space as the first random variable time series. Second random variable time series ; Introducing time delay factor The time series of the second random variable is time-shifted to construct a time-shifted time series. Calculate the time series of the first random variable With the time-shifted time series The joint probability distribution density between them and their respective marginal probability distribution densities; Based on the principle of information entropy, the time series of the first random variable is calculated. With the time-shifted time series time delay mutual information value The maximum time delay mutual information value among the time delay mutual information values ​​corresponding to each time delay factor is defined as the dynamic nonlinear coupling coefficient between the two corresponding parameter nodes. According to different deep peak shaving conditions of thermal power plants, the preset association threshold is dynamically adjusted. When the dynamic nonlinear coupling coefficient under a certain deep peak shaving condition is lower than the preset association threshold, the corresponding edge in the dynamic nonlinear state topology graph is erased to realize the adaptive variable structure dimensionality reduction of the topology.

[0007] In some embodiments, the step of decomposing the dynamic nonlinear state topology diagram into multiple collaboratively related digital state cell units based on the topological boundaries and thermodynamic functional partitioning of the physical equipment in a thermal power plant specifically includes: Based on the principle of system-level decoupling, the overall topology of the entire plant is divided into boiler superheater state cell units, turbine high-pressure cylinder state cell units, and generator stator state cell units. Each of the digital state cell units is configured with an independent virtual input boundary interface and a virtual output boundary interface. Adjacent digital state cell units can transmit mass flow data, energy flow data, and information flow data nonlinearly and dynamically through the virtual input boundary interface and the virtual output boundary interface to achieve distributed collaborative computing within the state cell matrix.

[0008] In some embodiments, the calculation of the local entropy growth rate and cumulative microscopic thermal stress fatigue damage degree of each digitized state cell unit based on a physical information neural network, and the output of a continuous spatiotemporal autocorrelation health evolution trajectory, specifically includes: Construct an equivalent three-dimensional spatiotemporal finite element mesh model for each of the digitized state cell units; The local microscopic thermodynamic parameters and material stress parameters extracted from the distributed monitoring stream data and after noise reduction are used as dynamic boundary conditions and input into the physical information neural network. The physical loss function constructed based on the laws of conservation of mass, momentum and energy is used for regularization constraints to calculate the non-uniform temperature field distribution and thermal stress field distribution on the surface and inside of the digitized state cell unit in three-dimensional space. By combining the rainflow counting method to count the number of alternating thermal stress cycles in the thermal stress field distribution, the cumulative fatigue damage coefficient is calculated, and the predicted projection line of the cumulative fatigue damage coefficient on the future time axis is defined as the spatiotemporal autocorrelation health evolution trajectory.

[0009] In some embodiments, the multi-scenario forward-looking evolutionary simulation specifically includes: Extract the rate of change and step amplitude of the target load command demand to be executed, and automatically match and generate multi-scenario simulation sequences including extreme load increase scenario, deep steady-state peak shaving scenario and rapid load reduction scenario; Multiple independent simulation threads run in parallel within the spatiotemporal collaborative digital twin evolution field. Each simulation thread loads a scenario simulation sequence, predicts the nonlinear response of each digitized state cell unit in the state cell matrix under future variable operating condition disturbances, and monitors in real time whether each parameter node approaches the limit safety boundary in the damage critical threshold matrix, so as to output the device safety probability distribution index under each scenario.

[0010] In some embodiments, S6 specifically includes: A multi-objective genetic algorithm based on Pareto optimality is adopted to establish a first objective function with minimizing the standard coal consumption of thermal power plants as the first objective, and a second objective function with minimizing the total cumulative fatigue damage increment of the state cell matrix as the second objective. Under the premise of satisfying the overall main steam pressure safety constraint boundary, the main steam temperature safety constraint boundary, and the grid peak-shaving response delay boundary of the thermal power plant, the optimal Pareto solution set that takes into account both the overall plant operating energy efficiency and the long-term equipment life is calculated. The set of solutions with the highest harmonic evaluation index is selected from the optimal Pareto solution set as the optimal load allocation matrix, and the boiler feedwater flow fine-tuning amount, turbine control valve opening fine-tuning amount, and forced draft matching value associated with the optimal load allocation matrix are converted into the dynamic control fine-tuning parameter stream.

[0011] In some embodiments, after the optimal load allocation matrix and the dynamic control fine-tuning parameter stream are sent to the distributed control system for execution, the method further includes: S7. Real-time acquisition of actual physical data after the response of the distributed control system. Comparison of the actual physical data with the simulation prediction data of the spatiotemporal collaborative digital twin evolution field at the corresponding time point to calculate the real-time residual vector. When the magnitude of the real-time residual vector is greater than the preset error tolerance threshold, it is determined that the current spatiotemporal collaborative digital twin evolution field has operational drift, and a state update mechanism is triggered. The state update mechanism uses the extended Kalman filter algorithm to perform online state estimation and correction of the dynamic nonlinear coupling coefficient in S2, and performs online recalibration of the dynamic nonlinear state topology until the magnitude of the updated real-time residual vector converges to within the error tolerance threshold.

[0012] This invention also provides a digital-based thermal power plant status data processing system, comprising: An adaptive topology construction module is used to acquire distributed monitoring flow data of the entire process of a thermal power plant in real time to construct an original digital state space containing multiple parameter nodes, and calculate the dynamic nonlinear coupling coefficient between any two parameter nodes in the original digital state space based on the time-delay mutual information algorithm. According to the dynamic nonlinear coupling coefficient, the edges that do not meet the preset association threshold are pruned and dimension-reduced to generate a dynamic nonlinear state topology graph that characterizes the real-time operating conditions of the thermal power plant. The healthy cell evolution matrix module is used to decompose the dynamic nonlinear state topology map into multiple collaboratively related digital state cell units to form a state cell matrix based on the topological boundaries and thermodynamic functions of the physical equipment in the thermal power plant. For each digital state cell unit, local microscopic thermodynamic parameters, vibration spectrum parameters, and material stress parameters are extracted independently. Based on the physical information neural network, the local entropy growth rate and the cumulative damage degree of microscopic thermal stress fatigue of each digital state cell unit are calculated, and a continuous spatiotemporal autocorrelation health evolution trajectory is output to establish a damage critical threshold matrix for each digital state cell unit. The digital twin evolution field simulation module is used to receive the target load command requirements to be executed from the power grid end, and input the target load command requirements to be executed as boundary conditions into the spatiotemporal collaborative digital twin evolution field composed of the state cell matrix to carry out multi-scenario forward-looking evolution simulation, and simulate the nonlinear state transition process of the state cell matrix in the future time window. The plant-wide collaborative load optimization terminal is used to dynamically calculate the optimal load allocation matrix and dynamic control fine-tuning parameter flow of each subsystem of the plant under the premise of not exceeding the damage critical threshold of any of the digital state cell units during the forward-looking evolution simulation of the multi-scenario scenario. The terminal aims to optimize the plant-wide fuel economy and maximize the health and lifespan of each digital state cell unit as the collaborative control objectives. The terminal then sends the optimal load allocation matrix and the dynamic control fine-tuning parameter flow to the distributed control system for execution.

[0013] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the digital-based thermal power plant status data processing method described above.

[0014] The present invention also provides a digital data processing device for thermal power plants, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the digital thermal power plant status data processing method described above.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. By constructing the original digital state space of multi-parameter nodes and extracting time-delay mutual information, adaptive topological dimensionality reduction of massive high-dimensional streaming data from thermal power plants is achieved. This not only significantly reduces the computational load on the central processing unit, but also helps to preserve the essential nonlinear coupling characteristics and transient operating condition variations between heterogeneous physical parameters, and significantly reduces the data distortion caused by the forced linearization tuning of traditional schemes.

[0016] 2. By introducing a digital state cell matrix that deeply decouples space and function, macroscopic devices are "decoupled" into microscopic white box units. A physical information neural network (PINN) that forcibly integrates the conservation laws of fluid mechanics and thermodynamics is loaded, which can output the internal stress fatigue and local entropy increase trajectory of the component surface and interior in real time. This effectively solves the pain points of poor extrapolation prediction and easy violation of physical red lines of pure data-driven neural networks.

[0017] 3. By constructing a spatiotemporally coordinated digital twin evolution field to conduct forward-looking parallel simulations of future load demands across multiple scenarios, and combining this with Pareto multi-objective genetic algorithm optimization, the control decision-making is transformed from the traditional "post-event passive maintenance and replacement" to "pre-event proactive life-cycle energy efficiency coordinated regulation." The algorithm can be used to dynamically delay the fatigue damage rate of core physical hardware online, providing strong safety and economic guarantees for the frequent peak-shaving operation of thermal power plants under the high proportion of new energy grid connection. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the overall process of a digital-based thermal power plant status data processing method in an embodiment of the present invention.

[0019] Figure 2 This is a schematic diagram illustrating the principle of using the time-delay mutual information algorithm to solve the nonlinear coupling coefficient and construct the dynamic topology graph in step S2 of this embodiment of the invention.

[0020] Figure 3 This is a block diagram showing the core module composition and interaction structure of the digital thermal power plant status data processing system in this embodiment of the invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] It should be noted that the methods and systems provided in this application can be deployed on advanced computing servers, edge computing nodes, or cloud data centers above the plant-level information monitoring system (SIS) and distributed control system (DCS) of a thermal power plant. The processing objects include high-frequency, multi-dimensional industrial real-time streaming data generated by tens of thousands of sensors during the operation of supercritical and ultra-supercritical coal-fired power generating units.

[0023] The specific implementation methods provided in this application will be described in detail below with reference to the accompanying drawings.

[0024] like Figure 1 As shown, the digital-based thermal power plant status data processing method specifically includes the following steps: S1, acquiring distributed monitoring stream data of the entire thermal power plant process in real time, and constructing an original digital state space containing multiple parameter nodes. A "parameter node" refers to a sensor measurement point or a state variable obtained by fusing multiple measurement points. In this embodiment, the data processing device communicates with the underlying DCS or SIS system of the thermal power plant at high concurrency through standard industrial network protocols (e.g., OPCUA, ModbusTCP, Profinet, or MQTT), acquiring the distributed monitoring stream data of the entire process in real time at a preset sampling frequency (e.g., 1Hz for thermodynamic parameters such as temperature and pressure, and 50Hz-100Hz for mechanical dynamic parameters such as rotor vibration and displacement).

[0025] Due to the strong electromagnetic interference, severe equipment vibration, and high temperature and pressure environment at the physical production site of thermal power plants, the raw monitoring stream data inevitably contains measurement noise and outliers caused by disconnections. To ensure the accuracy of subsequent feature extraction, in this embodiment, the data processing device first calls the sliding median filter algorithm to remove spike pulse outliers in the discrete data stream, and then uses a first-order low-pass Butterworth digital filter for smoothing and noise reduction.

[0026] Because the various measuring points in a thermal power plant cover temperature... Pressure (MPa), flow rate (t / h), vibration amplitude The data includes physical quantities with various dimensions, such as electrical power (MW), resulting in significant differences in absolute values. To eliminate the influence of dimensions on subsequent nonlinear correlation calculations, the data processing equipment performs normalization or Z-score standardization preprocessing on the smoothed and denoised time series of each measurement point. After preprocessing, the system constructs an M-dimensional original digital state space in the computing memory, mapping each standardized measurement point parameter (such as boiler main steam temperature, main steam pressure, reheat steam temperature, turbine control valve opening, high-pressure cylinder inner wall temperature, rotor shaft vibration spectrum, generator output active power, etc.) to an independent orthogonal axis node in this space. This original digital state space can completely characterize the digital operating state of the thermal power plant at any given sampling instant in the form of high-dimensional feature vectors.

[0027] S2, calculate the dynamic nonlinear coupling coefficient between any two parameter nodes in the original digital state space based on the time-delay mutual information algorithm, and prune and reduce the dimensionality of edges that do not meet the preset association threshold according to the dynamic nonlinear coupling coefficient, so as to generate a dynamic nonlinear state topology graph that represents the real-time operating conditions of the thermal power plant.

[0028] The thermodynamic processes within a thermal power plant are typical dynamic systems characterized by large time delays, strong coupling, and high nonlinearity. Traditional linear data compression schemes often forcibly set a single master parameter and align other parameters linearly with it, which leads to the loss of a large amount of characteristic information representing local transient variations or highly nonlinear operating conditions during the preprocessing stage.

[0029] To address this issue, embodiments of this application introduce a time-delay mutual information algorithm to unbiasedly extract the dynamic nonlinear coupling characteristics between any two heterogeneous parameter nodes. Figure 2 This is a schematic diagram illustrating the principle of using the time-delay mutual information algorithm to solve the nonlinear coupling coefficients and construct a dynamic topology graph in step S2 of this application embodiment. For example... Figure 2 As shown, the specific solution process includes: extracting two parameter nodes from the original digital state space in a pairwise pairwise manner, and defining their corresponding standardized discrete-time series as the first random variable time series. Second random variable time series .

[0030] Considering the inherent physical delay (phase lag) in the transfer of heat or working fluid through the complex boiler piping and turbine cylinder, embodiments of this application introduce a time delay factor. .

[0031] Data processing equipment for the second random variable time series A time-shifted time series is constructed by shifting the time axis to the right. .

[0032] Subsequently, using the kernel density estimation algorithm, the time series of the first random variable within the current scrolling window is statistically analyzed and calculated online. With time-shifted time series Joint probability distribution density between and the time series of the first random variable Its own marginal probability distribution density and time-shifted time series Its own marginal probability distribution density At this point, the data processing device calls the core calculation formula (1) to calculate the current time delay factor. Time delay mutual information value between the next two nodes .

[0033] ; In the above formula (1), the data processing device is within the preset maximum time delay span. Incremental traversal of time delay factors Since the time delay factor at the maximum value precisely corresponds to the actual physical time delay of thermodynamic conduction between two physical measurement points, the system extracts the maximum value on this spectral line, i.e., the maximum time delay mutual information value. And define this maximum value as the parameter node. With parameter nodes The dynamic nonlinear coupling coefficient between them.

[0034] When all of space After all the dynamic nonlinear coupling coefficients for each node are calculated, they are combined in the computing memory to form a dense equivalent adjacency matrix. To eliminate weak environmental clutter interference and significantly reduce the computational load of subsequent parallel digital twin simulations, the system initiates an adaptive variable structure pruning mechanism. This mechanism dynamically calls a preset association threshold based on the current macroscopic peak-shaving conditions of the thermal power plant. The system compares the elements of the adjacency matrix one by one. If the dynamic nonlinear coupling coefficient between two parameter nodes is determined to be lower than the preset association threshold, the corresponding matrix element is cleared to zero, and the connection edge between the two nodes is removed at the topology level. Conversely, if the coupling coefficient is greater than or equal to the preset association threshold, the connection edge is fully retained, and the nonlinear coupling coefficient is directly used as the topological weight of the edge. After traversal, a sparse, adaptive variable structure dynamic nonlinear state topology graph is automatically constructed, achieving high-fidelity nonlinear dimensionality reduction of massive streaming data.

[0035] S3. Based on the topological boundaries and thermodynamic functional division of the physical equipment in the thermal power plant, the dynamic nonlinear state topology diagram is decomposed into multiple collaboratively related digital state cell units to form a state cell matrix.

[0036] To prevent the entire thermal power plant or steam turbine from being monitored in a crude manner as a single "black box" that cannot perceive the degradation of its internal structure, this application proposes a "cellular" matrix modeling method that deeply decouples system-level functions from physical boundaries.

[0037] Specifically, the data processing equipment uses a subgraph cutting algorithm to structurally partition the dynamic nonlinear state topology generated in step S2 based on the inherent thermodynamic mass conservation boundary and energy isolation interface in the physical thermal power plant. After partitioning, multiple functionally independent but deeply collaborative topology subgraphs are formed, which are defined as digital state cell units in this embodiment.

[0038] In this embodiment, the overall plant topology is standardized, decomposed, and reconstructed into at least three core cellular units: Boiler superheater state cell unit: It aggregates and encapsulates topological subgraph nodes that characterize the heat release from coal combustion on the boiler side, the temperature distribution of the superheater tube wall on the heating surface, and the enthalpy of the main steam outlet. High-pressure cylinder state cell unit of steam turbine: It aggregates and encapsulates topological subgraph nodes that characterize transient pressure after high-pressure main steam valve, extraction steam temperature of each section of cylinder, blade root stress, and three-dimensional vibration spectrum of rotor bearing; Generator stator state cell unit: It aggregates and encapsulates topological subgraph nodes that characterize generator speed, stator winding temperature, excitation current, and final output instantaneous power.

[0039] The aforementioned digitized state cell units are arranged together to form the plant's entire digitized state cell matrix. To ensure high-fidelity co-simulation within the matrix, the system configures standardized virtual input boundary interfaces and virtual output boundary interfaces for each state cell unit in its software architecture. Adjacent state cell units communicate with each other via these virtual input and output boundary interfaces, using high-dimensional feature tensors to perform real-time nonlinear dynamic coupling and transmission of mass flow data (such as working fluid flow rate), energy flow data (such as sensible enthalpy and thermal stress energy), and information flow data, thereby achieving distributed collaborative computing within the state cell matrix.

[0040] S4. For each digitized state cell unit, local microscopic thermodynamic parameters, vibration spectrum parameters, and material stress parameters are independently extracted. Based on a physical information neural network, the local entropy growth rate and the cumulative damage degree of microscopic thermal stress fatigue for each digitized state cell unit are calculated, and a continuous spatiotemporal autocorrelation health evolution trajectory is output to establish a damage critical threshold matrix for each digitized state cell unit (the damage critical threshold matrix is ​​a matrix composed of thresholds such as material fatigue limit, upper temperature limit, and upper stress limit). After decoupling the state cell matrix, it is necessary to perform white-box microscopic fatigue damage and degradation trajectory calculation within each cell unit. Ordinary neural networks, when faced with extrapolation intervals or complex working conditions, are prone to data drift that violates physical laws such as energy conservation when the output results are not constrained by physical laws. Therefore, this embodiment of the application loads a physical information neural network (i.e., PINN) into the deep calculation of state cells.

[0041] Specifically, the data processing device first constructs an equivalent three-dimensional spatiotemporal finite element mesh model corresponding to each digitized state cell unit in the computing memory. Extracted and pre-processed with damped noise reduction, the local microscopic thermodynamic parameters and material stress parameters are input into the network as dynamic boundary conditions. In this embodiment, the fundamental physical conservation laws of thermodynamics and fluid mechanics are explicitly embedded as regularization penalty terms into the network's total training loss function.

[0042] ; In the above formula (2), the definitions of each mathematical and physical symbol are as follows: This represents the overall loss function value of the PINN neural network at the current training iteration step; This represents the mean square error loss between the microscopic physical quantity predicted by the neural network and the actual physical data collected by the sensor at the measurement point. These represent the regularization weight adjustment coefficients for the data loss term and the physical residual penalty term, respectively; This represents the dynamic density of the moving working fluid (such as ultra-supercritical water vapor) within the cell unit; This represents the instantaneous velocity vector of the working fluid. For the three-dimensional Hamiltonian operator, Represents divergence operations; Representing the continuous time dimension, The total derivative of a substance in thermodynamics (the piecewise total derivative terms of a substance are denoted as follows) ); Indicates the local instantaneous static pressure of a fluid. The viscous stress tensor represents the fluid's viscosity. This indicates the specific constant-pressure specific heat capacity of the metal material or working fluid of the monitored core component; It represents the spatiotemporal dynamic temperature field distribution on or inside the metal surface of a component; This indicates the inherent thermal conductivity of the core monitoring hardware material; It represents the intensity of the heat source per unit volume due to friction, eddy currents, or internal strain energy dissipation.

[0043] Through the explicit constraints of formula (2), the network weight updates are forcibly restricted within the physical redline, enabling real-time online accurate calculation of the non-uniform three-dimensional temperature and thermal stress field distributions within each state cell unit. Based on this, the system combines rainflow counting to statistically analyze the number of cycles and stress amplitude of alternating thermal stress in real time. Based on the linear damage accumulation theory, the system calculates the cumulative fatigue damage coefficient for each sampling step online. The predicted projection line of this damage coefficient on the future time axis is defined as the spatiotemporal autocorrelation health evolution trajectory. Simultaneously, a damage critical threshold matrix is ​​automatically established by combining the physical limits of the materials of each component.

[0044] S5, receive the target load instruction requirements to be executed from the power grid, and input the target load instruction requirements to be executed as boundary conditions into the spatiotemporal collaborative digital twin evolution field composed of the state cell matrix to perform multi-scenario forward-looking evolution simulation, and simulate the nonlinear state transition process of the state cell matrix in the future time window.

[0045] The embodiments of this application effectively break away from the traditional technology's reactive maintenance model of passively waiting for hardware degradation and damage before making replacement decisions, and instead construct a proactive and forward-looking virtual simulation architecture.

[0046] Specifically, the communication interface of the data processing equipment receives in real time the target load command requirements to be executed within the future variable operating condition time window issued by the external power grid dispatch center (e.g., requiring the unit to rapidly reduce the unit load to deep peak shaving condition at a drastic variable operating condition rate of 20MW per minute within the next 20 minutes). The system automatically extracts the rate of change, step amplitude, and duration characteristics of the target load command requirements to be executed.

[0047] Subsequently, the system activates a virtual "spatiotemporal collaborative digital twin evolution field" composed of state cell matrices at the software level, leveraging the multi-core advantage of the computing device's central processing unit to launch multiple independent virtual simulation threads in parallel. Each simulation thread independently loads a specific multi-scenario simulation sequence under varying operating conditions (e.g., extreme load increase scenario, deep steady-state peak shaving scenario, and rapid load reduction scenario). Within the virtual twin sandbox, each thread performs a fast-forward simulation along the time axis (e.g., predicting the response of each subsystem within the next 30 minutes), predicting the nonlinear state transition process of each digitized state cell unit in the state cell matrix under future varying operating condition disturbances, and monitoring in real time whether each parameter node approaches the limit safety boundary in the damage critical threshold matrix, thereby calculating and outputting the device safety probability distribution index under each forward-looking scenario.

[0048] S6, during the multi-scenario forward-looking evolution simulation, with the optimal fuel economy of the entire plant and the maximization of the health and lifespan of each of the digital state cell units as the collaborative control objectives, the optimal load allocation matrix and dynamic control fine-tuning parameter flow of each subsystem of the entire plant are dynamically calculated without exceeding the damage critical threshold of any of the digital state cell units, and the optimal load allocation matrix and the dynamic control fine-tuning parameter flow are sent to the distributed control system for execution.

[0049] In the process of forward-looking simulation of digital twin evolution field, the system transforms passive into active, calls the core calculation formula (3), and initiates a multi-objective genetic optimization algorithm based on Pareto optimal solution to actively find a set of comprehensive optimal control strategies at the software level that can maximize the energy efficiency of thermal power plants and the long lifespan of core hardware.

[0050] ; In the above main decision optimization formula (3), the definitions of each control variable and the multi-objective function are fully described as follows: This refers to the control vector space defined in the embodiments of this application, which contains dynamic control action variables of multiple subsystems throughout the plant. Specifically, It is a high-dimensional real number vector whose components explicitly include: boiler feedwater pump flow setpoint, coal mill speed, forced draft fan damper opening, and dynamic opening fine adjustment of each steam turbine steam distribution regulating valve; This represents the optimal Pareto control decision solution set calculated through global optimization. All solutions within this set collectively constitute an energy efficiency-lifetime frontier boundary line. Let be the first fuel economy objective function, representing the instantaneous standard coal consumption of the entire plant under the current control vector. The sum; Let be the second long-cycle fatigue life degradation objective function, representing the state of all cells in the plant-wide state cell matrix under the control vector. The incremental microscopic cumulative fatigue damage caused by future thermal stress disturbances The sum of .

[0051] The entire multi-objective optimization process is subject to extremely stringent industrial-grade boundary conditions. The optimization search space is strictly limited to: the overall main steam pressure of the thermal power plant must not exceed the material safety limit, and the main steam temperature fluctuation range must be maintained within the design range. Within the constraints and under the premise of the power grid peak-shaving response delay boundary, the predicted damage degree of any digital state cell must absolutely not exceed its damage critical threshold matrix.

[0052] After the algorithm iteration is completed, the system selects the solution with the highest comprehensive evaluation index as the optimal load allocation matrix, and transforms the bias value associated with this matrix into a dynamic control fine-tuning parameter stream. This parameter stream is directly sent to the DCS central execution controller at the bottom layer of the thermal power plant via the plant's industrial Ethernet, commanding the physical hardware devices to execute this proactive collaborative control decision.

[0053] S7. Real-time acquisition of actual physical feedback data after the response of the distributed control system. Comparison of the actual physical feedback data with the simulation prediction data of the spatiotemporal collaborative digital twin evolution field at the corresponding time point to calculate the real-time residual vector. When the magnitude of the real-time residual vector is greater than the preset error tolerance threshold, it is determined that the current spatiotemporal collaborative digital twin evolution field has operational drift, and the state update mechanism is triggered.

[0054] To prevent inevitable model drift between the digital calculation model and the actual physical power plant due to long-term operation (e.g., changes in pipeline heat transfer coefficient due to ash accumulation during operation, or sudden changes in fuel calorific value due to batch replacement of coal), this application embodiment further configures a highly dynamic self-calibration feedback update mechanism.

[0055] Specifically, in the next sampling period after the optimal control decision parameters are sent to the DCS for execution, the online residual checker in the system collects the actual physical feedback data vector returned by the DCS sensor response in real time, and compares it with the simulation prediction data vector of the digital twin evolution field at the corresponding time point to calculate the real-time residual vector.

[0056] The system continuously monitors the magnitude of the real-time residual vector. Once the magnitude of the residual vector exceeds the system's preset error tolerance threshold, the system automatically determines that a nonlinear operational drift has occurred between the evolution field and the physical entity, and immediately activates the online state recalibration and update mechanism. This update mechanism uses the Extended Kalman Filter (EKF) algorithm, taking the dynamic nonlinear coupling coefficients (i.e., the weights of the topology graph edges) from step S2 as the hidden state variables to be estimated, and using the latest real-time residual vector as the new weight gain input, it calculates and corrects each nonlinear coupling coefficient online, performs online recalibration and variable structure edge weight refresh on the dynamic nonlinear state topology graph, forcibly causing the magnitude of the real-time residual vector to converge again and fall within the preset error tolerance threshold, thus realizing the closed-loop fully autonomous adaptive evolution of the system.

[0057] The calculations involved in the embodiments of the present invention are all dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0058] In order to implement comprehensive protection at the industrial product level using the aforementioned advanced data processing methods, Figure 3 This document illustrates the core module composition and interaction structure of a digital-based thermal power plant status data processing system according to an embodiment of this application. The system adopts a high-concurrency, low-latency industrial IoT microservice architecture and is independently deployed within the computer equipment of the central digital computing cluster in the thermal power plant.

[0059] like Figure 3 As shown, this digital-based thermal power plant status data processing system specifically comprises the following four interconnected hardware and software modules: 1. Adaptive Topology Building Module This module is equipped with a standard industrial communication bus hardware interface card for executing steps S1 and S2. It acquires DCS / SIS distributed monitoring stream data from the entire thermal power plant process in real time and in parallel, constructing a multi-dimensional original digital state space in memory. The module integrates a dedicated time-delay mutual information acceleration operator, capable of efficiently solving the time-delay mutual information of discrete sequences pairwise (Equation 1), accurately capturing the dynamic nonlinear coupling coefficients between parameter nodes. Simultaneously, the module incorporates a built-in adaptive pruning mechanism that dynamically invokes preset association thresholds based on different operating conditions such as steady state and deep peak shaving to prune dense matrices, outputting a low-computation, high-fidelity dynamic nonlinear state topology diagram.

[0060] 2. Healthy Cell Evolution Matrix Module This module is seamlessly coupled to the data output of the adaptive topology construction module and is specifically used to execute steps S3 and S4. Internally, this module stores physical three-dimensional mesh topology boundary data for all plant equipment. It is responsible for logically cutting the input nonlinear topology graph into decoupled digital state cell units such as boiler superheaters, turbine high-pressure cylinders, and generator stators, each equipped with a virtual boundary interface, and then assembling them into a state cell matrix. The core hardware and software kernel of this module is a set of GPU-accelerated and optimized Physical Information Neural Network (PINN) inference engines. This engine explicitly constructs the mass, momentum, and energy conservation partial differential equations corresponding to formula (2) in the loss function. By reading local microscopic parameters in real time, the module can calculate the high-precision temperature field and stress field inside each cell unit online in real time, and, combined with the rainflow counting method, extrapolate and predict the spatiotemporal autocorrelation health evolution trajectory of each component and the full-dimensional damage critical threshold matrix.

[0061] 3. Digital Twin Evolution Field Simulation Module The input terminals of this module are coupled to the output terminals of the power grid dispatch interface and the healthy cell evolution matrix module, respectively, and are used to execute step S5. This module includes a multi-threaded parallel computing engine specifically responsible for receiving future target load command requirements from the power grid. The module creates an independent virtual simulation sandbox in memory, using future load commands as transient boundary conditions as input. It then uses multi-threaded parallel simulation to advance the nonlinear state transition and diffusion process of the state cell matrix within future time windows under multiple sub-scenarios such as extreme load increases and rapid load decreases, outputting in real-time the probability distribution indicators of the future safety of the entire plant's core hardware.

[0062] 4. Plant-wide collaborative load optimization terminal As the decision-making and control brain of the entire system, the terminal's input end is tightly coupled to the scene output end of the digital twin evolution field simulation module, while its control signal output end is physically connected to the central control unit of the distributed control system (DCS) in the plant through a shielded industrial network cable, and is used to execute the steps S6 and S7. The terminal integrates a multi-objective genetic optimization operator based on Pareto optimal solution, with formula (3) as the core fitness function. Under the premise of satisfying the safety boundary constraints such as main steam temperature and pressure, the global optimization solution calculates the optimal load allocation matrix with dual optimal energy efficiency and equipment life, which is then converted into a dynamic control fine-tuning parameter stream and sent to the DCS bottom loop for execution in real time.

[0063] Meanwhile, the plant-wide collaborative load optimization terminal is equipped with an online residual checker and an online extended Kalman filter (EKF) calibrator. The checker continuously compares the actual physical data returned by the DCS with the predicted data of the twin field. Once the residual exceeds the limit, the EKF calibration operator is immediately triggered to calculate the online correction amount of the nonlinear coupling coefficient. The calibration signal is then forcibly input into the adaptive topology construction module through a reverse feedback calibration bus to perform online reconstruction calibration of the dynamic nonlinear state topology diagram, causing the real-time residual vector to converge again and fall within the preset error tolerance threshold.

[0064] This application also provides a computer-readable storage medium, such as non-volatile dynamic random access memory (NVRAM), flash memory, solid-state drive (SSD), or optical disk storage. The storage medium stores computer program code instructions. When the storage medium is read and executed by the processor of a digital computing device in a thermal power plant, the processor will strictly follow the method steps S1 to S7 described in the above method embodiments to drive the coordinated operation of various hardware modules, completing nonlinear dimensionality reduction, cellular decoupling, physical conservation trajectory prediction, twin simulation, and multi-objective Pareto optimization decision output for energy efficiency and lifetime of massive distributed streaming data.

[0065] This application also provides a digital data processing device for thermal power plants, the hardware of which is deployed in the central control room of the power plant. The data processing device includes at least: a set of high-performance multi-core central processing units (CPUs) symmetrically coupled via a high-speed interconnect bus; a set of graphics processing units (GPUs) dedicated to parallel computation of tensors and neural networks; a set of large-capacity high-speed system memory (RAM); and a standard industrial Ethernet network adapter. The memory stores executable computer programs. When the processor calls and runs the computer programs, the processing device is fully activated and transformed into the adaptive topology construction module, healthy cell evolution matrix module, digital twin evolution field deduction module, and plant-wide collaborative load optimization terminal described in this application's embodiment, thereby comprehensively realizing the entire process flow of the digital-based thermal power plant status data processing method.

[0066] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wireless or wired transmission; wired transmission methods include optical fiber, twisted pair, coaxial cable, etc.; wireless transmission includes infrared, microwave, etc. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center containing one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0067] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0068] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0069] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0070] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0071] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion 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 this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.

[0072] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for processing digitalized thermal power plant status data, characterized in that, Includes the following steps: S1 acquires distributed monitoring stream data of the entire process of thermal power plants in real time and constructs an original digital state space containing multiple parameter nodes. S2, calculate the dynamic nonlinear coupling coefficient between any two parameter nodes in the original digital state space based on the time-delay mutual information algorithm, and prune and reduce the dimensions of edges that do not meet the preset association threshold according to the dynamic nonlinear coupling coefficient, so as to generate a dynamic nonlinear state topology graph that characterizes the real-time operating conditions of the thermal power plant. S3, based on the topological boundaries and thermodynamic functional division of the physical equipment of the thermal power plant, the dynamic nonlinear state topology diagram is decomposed into multiple collaboratively related digital state cell units to form a state cell matrix. S4. For each of the digital state cell units, the local microscopic thermodynamic parameters, vibration spectrum parameters and material stress parameters are extracted independently. Based on the physical information neural network, the local entropy growth rate and the cumulative damage degree of microscopic thermal stress fatigue of each of the digital state cell units are calculated, and a continuous spatiotemporal autocorrelation health evolution trajectory is output to establish the damage critical threshold matrix of each of the digital state cell units. S5, receive the target load instruction requirements to be executed from the power grid, and input the target load instruction requirements to be executed as boundary conditions into the spatiotemporal collaborative digital twin evolution field composed of the state cell matrix to perform multi-scenario forward-looking evolution simulation, and simulate the nonlinear state transition process of the state cell matrix in the future time window. S6, during the multi-scenario forward-looking evolution simulation, with the optimal fuel economy of the entire plant and the maximization of the health and lifespan of each of the digital state cell units as the collaborative control objectives, the optimal load allocation matrix and dynamic control fine-tuning parameter flow of each subsystem of the entire plant are dynamically calculated without exceeding the damage critical threshold of any of the digital state cell units, and the optimal load allocation matrix and the dynamic control fine-tuning parameter flow are sent to the distributed control system for execution.

2. The method for processing digitalized thermal power plant status data according to claim 1, characterized in that, S2 specifically includes: Define the discrete time series corresponding to any two parameter nodes in the original digital state space as the first random variable time series. Second random variable time series ; Introducing time delay factor The time series of the second random variable is time-shifted to construct a time-shifted time series. Calculate the time series of the first random variable With the time-shifted time series The joint probability distribution density between them and their respective marginal probability distribution densities; Based on the principle of information entropy, the time series of the first random variable is calculated. With the time-shifted time series time delay mutual information value The maximum time delay mutual information value among the time delay mutual information values ​​corresponding to each time delay factor is defined as the dynamic nonlinear coupling coefficient between the two corresponding parameter nodes. According to different deep peak shaving conditions of thermal power plants, the preset association threshold is dynamically adjusted. When the dynamic nonlinear coupling coefficient under a certain deep peak shaving condition is lower than the preset association threshold, the corresponding edge in the dynamic nonlinear state topology graph is erased to realize the adaptive variable structure dimensionality reduction of the topology.

3. The method according to claim 1, characterized in that, Based on the topological boundaries and thermodynamic functional division of the physical equipment in the thermal power plant, the dynamic nonlinear state topology diagram is decomposed into multiple collaboratively related digital state cell units, specifically including: Based on the principle of system-level decoupling, the overall topology of the entire plant is divided into boiler superheater state cell units, turbine high-pressure cylinder state cell units, and generator stator state cell units. Each of the digital state cell units is configured with an independent virtual input boundary interface and a virtual output boundary interface. Adjacent digital state cell units can transmit mass flow data, energy flow data, and information flow data nonlinearly and dynamically through the virtual input boundary interface and the virtual output boundary interface to achieve distributed collaborative computing within the state cell matrix.

4. The method according to claim 1, characterized in that, The calculation of the local entropy growth rate and cumulative microscopic thermal stress fatigue damage degree of each digitized state cell unit based on the physical information neural network, and the output of a continuous spatiotemporal autocorrelation health evolution trajectory, specifically includes: Construct an equivalent three-dimensional spatiotemporal finite element mesh model for each of the digitized state cell units; The local microscopic thermodynamic parameters and material stress parameters extracted from the distributed monitoring stream data and after noise reduction are used as dynamic boundary conditions and input into the physical information neural network. The physical loss function constructed based on the laws of conservation of mass, momentum and energy is used for regularization constraints to calculate the non-uniform temperature field distribution and thermal stress field distribution on the surface and inside of the digitized state cell unit in three-dimensional space. By combining the rainflow counting method to count the number of alternating thermal stress cycles in the thermal stress field distribution, the cumulative fatigue damage coefficient is calculated, and the predicted projection line of the cumulative fatigue damage coefficient on the future time axis is defined as the spatiotemporal autocorrelation health evolution trajectory.

5. The method according to claim 1, characterized in that, The aforementioned multi-scenario forward-looking evolutionary simulation specifically includes: Extract the rate of change and step amplitude of the target load command demand to be executed, and automatically match and generate multi-scenario simulation sequences including extreme load increase scenario, deep steady-state peak shaving scenario and rapid load reduction scenario; Multiple independent simulation threads run in parallel within the spatiotemporal collaborative digital twin evolution field. Each simulation thread loads a scenario simulation sequence, predicts the nonlinear response of each digitized state cell unit in the state cell matrix under future variable operating condition disturbances, and monitors in real time whether each parameter node approaches the limit safety boundary in the damage critical threshold matrix, so as to output the device safety probability distribution index under each scenario.

6. The method according to claim 1, characterized in that, S6 specifically includes: A multi-objective genetic algorithm based on Pareto optimality is adopted to establish a first objective function with minimizing the standard coal consumption of thermal power plants as the first objective, and a second objective function with minimizing the total cumulative fatigue damage increment of the state cell matrix as the second objective. Under the premise of satisfying the overall main steam pressure safety constraint boundary, the main steam temperature safety constraint boundary, and the grid peak-shaving response delay boundary of the thermal power plant, the optimal Pareto solution set that takes into account both the overall plant operating energy efficiency and the long-term equipment life is calculated. The set of solutions with the highest harmonic evaluation index is selected from the optimal Pareto solution set as the optimal load allocation matrix, and the boiler feedwater flow fine-tuning amount, turbine control valve opening fine-tuning amount, and forced draft matching value associated with the optimal load allocation matrix are converted into the dynamic control fine-tuning parameter stream.

7. The method according to claim 1, characterized in that, After the optimal load allocation matrix and the dynamic control fine-tuning parameter stream are sent to the distributed control system for execution, the method further includes: S7. Real-time acquisition of actual physical data after the response of the distributed control system. Comparison of the actual physical data with the simulation prediction data of the spatiotemporal collaborative digital twin evolution field at the corresponding time point to calculate the real-time residual vector. When the magnitude of the real-time residual vector is greater than the preset error tolerance threshold, it is determined that the current spatiotemporal collaborative digital twin evolution field has operational drift, and a state update mechanism is triggered. The state update mechanism uses the extended Kalman filter algorithm to perform online state estimation and correction of the dynamic nonlinear coupling coefficient in S2, and performs online recalibration of the dynamic nonlinear state topology until the magnitude of the updated real-time residual vector converges to within the error tolerance threshold.

8. A digital-based thermal power plant status data processing system, characterized in that, include: An adaptive topology construction module is used to acquire distributed monitoring flow data of the entire process of a thermal power plant in real time to construct an original digital state space containing multiple parameter nodes, and calculate the dynamic nonlinear coupling coefficient between any two parameter nodes in the original digital state space based on the time-delay mutual information algorithm. According to the dynamic nonlinear coupling coefficient, the edges that do not meet the preset association threshold are pruned and dimension-reduced to generate a dynamic nonlinear state topology graph that characterizes the real-time operating conditions of the thermal power plant. The healthy cell evolution matrix module is used to decompose the dynamic nonlinear state topology map into multiple collaboratively related digital state cell units to form a state cell matrix based on the topological boundaries and thermodynamic functions of the physical equipment in the thermal power plant. For each digital state cell unit, local microscopic thermodynamic parameters, vibration spectrum parameters, and material stress parameters are extracted independently. Based on the physical information neural network, the local entropy growth rate and the cumulative damage degree of microscopic thermal stress fatigue of each digital state cell unit are calculated, and a continuous spatiotemporal autocorrelation health evolution trajectory is output to establish a damage critical threshold matrix for each digital state cell unit. The digital twin evolution field simulation module is used to receive the target load command requirements to be executed from the power grid end, and input the target load command requirements to be executed as boundary conditions into the spatiotemporal collaborative digital twin evolution field composed of the state cell matrix to carry out multi-scenario forward-looking evolution simulation, and simulate the nonlinear state transition process of the state cell matrix in the future time window. The plant-wide collaborative load optimization terminal is used to dynamically calculate the optimal load allocation matrix and dynamic control fine-tuning parameter flow of each subsystem of the plant under the premise of not exceeding the damage critical threshold of any of the digital state cell units during the forward-looking evolution simulation of the multi-scenario scenario. The terminal aims to optimize the plant-wide fuel economy and maximize the health and lifespan of each digital state cell unit as the collaborative control objectives. The terminal then sends the optimal load allocation matrix and the dynamic control fine-tuning parameter flow to the distributed control system for execution.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

10. A digital data processing device for a thermal power plant, comprising a memory and a processor, characterized in that, The memory is used to store computer programs, and the processor is used to call and run the computer programs stored in the memory to perform the method as described in any one of claims 1 to 7.

Images