Digital-twin-based dynamic optimization operation method, system and medium for gas-steam combined cycle unit

Abstract

The application relates to the technical field of generator set optimization control, and particularly discloses a gas-steam combined cycle unit dynamic optimization operation method, system and medium based on digital twinning. The system comprises a model management module, a state monitoring and decision module, a digital twinning simulation module and an optimization execution module. The core of the method is that the change rates of unit load instructions and key parameters are monitored in real time to determine the operation condition; based on the operation condition determination result, a matched digital twin is selected and assembled from the pre-established high-fidelity and fast two types of model clusters; and the high-fidelity twin is used to execute economic optimization for the steady-state condition, and the fast twin is used to execute rolling optimization for the transient condition. Through the dynamic self-reconstruction capability of the digital twin, the contradiction between model precision and calculation speed in the optimization process is solved, the adaptive optimization demand under different operation conditions is realized, and the economy, rapidity and stability of the unit operation are simultaneously improved.

Description

Technical Field

[0001] This invention relates to the field of generator set optimization control technology, and more specifically, to a method, system, and medium for dynamic optimization operation of gas-steam combined cycle units based on digital twins. Background Technology

[0002] Gas-fired combined cycle (Gas-Steam) units play a crucial role in peak shaving and baseload operations in modern power systems due to their high thermal efficiency and flexible operating characteristics. With the large-scale grid integration of renewable energy, the demands on the load response speed and operational economy of traditional generator units are increasing, causing these units to frequently operate under dynamic load variations. Against this backdrop, achieving dynamic optimization of unit operation across the entire operating range—that is, continuously pursuing optimal economic efficiency while ensuring rapid and stable dynamic processes—is of great significance.

[0003] Existing dynamic optimization schemes typically rely on pre-defined optimization algorithms and fixed unit simulation models. Among these, optimization methods based on digital twins, which predict and optimize by establishing virtual mappings of the unit, have become an important direction. However, most digital twins in existing technologies are single models with fixed structures. These models often face a contradiction during construction: to ensure optimization accuracy, high-fidelity models with complex mechanisms and detailed parameters are needed, but this leads to heavy computational loads and slow simulation speeds, making it difficult to meet the timeliness requirements of real-time optimization in rapid dynamic processes; if overly simplified models are used to meet real-time requirements, the accuracy and economy of the optimization results will significantly decrease under steady-state or slowly changing operating conditions. Furthermore, fixed-structure digital twins lack the ability to adapt to operating conditions and cannot dynamically adjust their simulation strategies based on the fundamentally different task requirements of steady-state economic optimization versus transient process tracking. This results in the system either running simple tasks at high computational cost, wasting resources, or missing optimization opportunities in dynamic processes due to insufficient model capabilities.

[0004] Therefore, the main problem with existing technologies is that static digital twin models are difficult to effectively reconcile the dual requirements of model accuracy and speed in dynamic optimization processes. Their fixed structure is mismatched with the dynamic and ever-changing optimization tasks, which limits the improvement of the overall performance of dynamic optimization. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method, system, and medium for dynamic optimization of operation of gas-steam combined cycle units based on digital twins, in order to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a dynamic optimization operation system for a gas-steam combined cycle unit based on digital twins, comprising:

[0007] The model management module serves as the foundation for storage and retrieval. It is used to classify and store simulation sub-models for gas turbine, waste heat boiler and steam turbine processes according to thermal processes. Each process stores at least one high-fidelity sub-model that focuses on simulation accuracy and one fast sub-model that focuses on calculation speed.

[0008] The status monitoring and decision-making module, as the perception and judgment center of the operating status, is used to collect unit operating data and external dispatch instructions in real time. It quantifies the intensity of dynamic processes by calculating the load instruction change rate and the key parameter deviation change rate, and compares the change rate with preset steady-state and transient thresholds to distinguish between stable and rapidly changing operating conditions. Based on the comparison results, it generates a first switching instruction indicating the use of a high-fidelity model combination or a second switching instruction indicating the use of a fast model combination.

[0009] The digital twin simulation module, as the core of model dynamic assembly and execution, is used to call and couple the corresponding high-fidelity sub-model from the model management module according to the first switching instruction, or to call and couple the corresponding fast sub-model according to the second switching instruction, so that a digital twin with different structures may be constructed each time the operating conditions change, and the twin is used to simulate and predict the future state of the unit.

[0010] The optimization execution module, acting as an optimization strategy matcher and solver, is used to perform steady-state optimization calculations with the goal of full-condition economic indicators based on the accurate prediction data provided by the high-fidelity sub-model when the first switching instruction is generated; and to perform rolling optimization calculations with the goal of dynamic process tracking performance based on the fast prediction data provided by the fast sub-model when the second switching instruction is generated.

[0011] Furthermore, the specific method by which the status monitoring and decision-making module generates instructions based on the comparison results is as follows:

[0012] S1: Real-time calculation of the load command change rate and the deviation change rate between the actual value and the set value of the main steam pressure at the current moment. These two change rates directly reflect the severity of changes in external command demand and internal state.

[0013] S2: If the load command change rate continues to be lower than the first threshold and the deviation change rate continues to be lower than the second threshold, it indicates that the unit is in the operating range of stable command and stable state, then the first switching command is generated to enable the high-fidelity model for fine simulation.

[0014] S3: If the load command change rate exceeds a higher third threshold or the deviation change rate exceeds a higher fourth threshold, it indicates that the unit is facing or about to enter a rapid dynamic process. Then, a second switching command is generated to enable the fast model to ensure the real-time performance of the simulation.

[0015] Furthermore, when executing step S3, the status monitoring and decision-making module also includes prediction logic: if it is determined that the load command change rate is about to exceed the third threshold based on the known future load plan curve or the short-term prediction based on the current command change trend, then a second switching command is generated in advance before the actual dynamic process of the unit begins, so as to reserve preparation time for optimized control.

[0016] Furthermore, the specific process by which the digital twin simulation module calls and couples the sub-model according to the switching command is as follows:

[0017] S1: Parse the switching instruction, query the preset mapping relationship according to the instruction type, and determine the set of high-fidelity or fast sub-model identifiers to be called;

[0018] S2: Based on the determined set of identifiers, load the corresponding sub-model program from the model management module into the runtime memory and complete the instantiation;

[0019] S3: Based on the actual energy flow and working fluid flow direction of the unit, the input and output interfaces of the loaded sub-models are physically connected in a consistent manner to form a complete system model that can be co-simulated.

[0020] Furthermore, the specific method by which the optimization execution module performs steady-state optimization is as follows:

[0021] S1: The comprehensive cost function, which includes fuel consumption costs and emission environmental indicators, is used as the optimization objective. This objective reflects the economic nature of steady-state operation.

[0022] S2: Based on the accurate prediction data obtained from simulating different steady-state setpoints using a high-fidelity digital twin, mathematical solutions are performed under the constraints of strictly meeting the safety operation boundary conditions of various equipment to obtain the steady-state setpoint that optimizes the overall cost.

[0023] Furthermore, the specific method by which the optimization execution module performs rolling optimization is as follows:

[0024] S1: The optimization objective is to minimize the prediction tracking error of key parameters (such as power and pressure) within a finite future time period. This objective aims to ensure the speed and stability of the dynamic process.

[0025] S2: At the beginning of each control cycle, based on the rapid prediction data of future dynamics from the fast digital twin, an optimization problem in the finite time domain is solved online, and the control command expected to be executed in the next week is immediately output to achieve closed-loop feedback control.

[0026] A method for dynamic optimization operation of a gas-steam combined cycle unit based on digital twins, comprising the following steps:

[0027] S100: Pre-build a model library to establish and store high-fidelity sub-models and fast sub-models of gas turbine, waste heat boiler and steam turbine processes, providing a basis for dynamic selection;

[0028] S200: Online monitoring and quantification, real-time data collection and computer group load command change rate and key parameter deviation change rate, transforming continuous operating status into identifiable characteristic quantities;

[0029] S300: Operation mode decision, compare the rate of change with preset steady-state and transient thresholds, and generate a first switching instruction or a second switching instruction accordingly, thereby deciding on the current digital twin model configuration to be adopted;

[0030] S400: Dynamic reconstruction of digital twins. Based on the generated switching instructions, it dynamically selects and couples the corresponding high-fidelity or fast sub-models, assembles them into a digital twin that matches the current working conditions, and performs real-time simulation prediction.

[0031] S500: Optimize strategy matching execution. If a first switching instruction is generated, then perform steady-state optimization with economy as the core based on high-fidelity simulation results; if a second switching instruction is generated, then perform rolling optimization with dynamic tracking performance as the core based on fast simulation results.

[0032] Furthermore, in step S300, the condition for generating the first switching command is that the load command change rate is continuously lower than the first threshold and the key parameter deviation change rate is continuously lower than the second threshold, which corresponds to the stable operation of the unit; the condition for generating the second switching command is that the load command change rate exceeds the higher third threshold or the key parameter deviation change rate exceeds the higher fourth threshold, which corresponds to the unit being in a rapidly changing transient operating condition.

[0033] Furthermore, the steady-state optimization described in step S500 specifically involves solving for and outputting the optimal steady-state setpoint under safety constraints, with the overall operating cost as the objective function; the rolling optimization specifically involves solving for and outputting the recent control command under process constraints, with the future dynamic tracking error as the objective function.

[0034] A computer-readable storage medium storing a computer program, which, when executed by a processor, controls the processor to sequentially execute the steps of the method described above, thereby realizing the dynamic optimization operation control logic of the gas-steam combined cycle unit at the hardware level.

[0035] The technical effects and advantages of this invention are as follows:

[0036] Compared to existing technologies that use fixed-structure digital twin models, this invention establishes a model cluster comprising two types of sub-models: high-fidelity and fast models, and designs a state monitoring and decision-making module. This module analyzes the rate of change of unit load commands and the rate of change of key parameter deviations in real time, automatically determining whether the current operating condition is steady-state or transient, and generating different model switching commands accordingly. The digital twin simulation module dynamically assembles the corresponding sub-model combinations based on the commands. This allows the system to use a high-fidelity model under steady-state conditions requiring detailed economic analysis, and a fast model under transient conditions requiring rapid response. Structurally, this resolves the contradiction between model accuracy and computational speed, ensuring that the optimization engine always possesses the most suitable predictive capability for the current task.

[0037] Specifically, when the unit faces transient processes such as rapid load increases or decreases, the system improves the simulation speed of the digital twin by activating a combination of fast sub-models with low computational overhead. This allows the subsequent rolling optimizer to solve for control variables at a higher frequency based on faster simulation predictions, thereby generating more timely optimization instructions that better reflect the actual evolution of the dynamic process. The effect is to enhance the system's ability to track rapid changes and improve dynamic control quality, reduce overshoot and fluctuations, and ensure the safety and stability of the unit during transient processes.

[0038] When the unit is running stably, the system automatically switches to a high-fidelity sub-model combination for simulation. The high-fidelity model provides more accurate predictions of consumption and emission characteristics, enabling the steady-state optimizer to perform more refined economic optimization and calculate a setpoint closer to the true optimal operating condition. This process avoids optimization biases caused by using simplified models, thus fully exploring the unit's energy-saving potential at different steady-state load points and improving long-term operational economic benefits.

[0039] Through the predictive logic in the state monitoring and decision-making module, the system can identify upcoming transient processes in advance based on known future load plans and switch to fast model configuration in advance. This predictive reconfiguration mechanism gives the optimizer valuable initialization and warm-up time, ensuring that the system is in optimal readiness when the dynamic process actually begins, eliminating control lag that may be caused by model switching and optimizer startup delays. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the overall system structure provided in an embodiment of the present invention.

[0041] Figure 2 This is a schematic diagram of the dynamic reconstruction process of a digital twin provided in an embodiment of the present invention.

[0042] Figure 3 This is a flowchart of the dynamic optimization operation closed loop provided in the embodiments of the present invention. Detailed Implementation

[0043] 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.

[0044] Example 1

[0045] As attached Figures 1 to 3 The specific implementation details of the dynamic optimization operation method, system, and medium for gas-steam combined cycle units based on digital twins are as follows:

[0046] Phase 1: System Preparation and Model Library Construction

[0047] This phase is completed before the system goes online and aims to establish the model foundation and strategy framework required for dynamic optimization.

[0048] S100: Construction and Storage of Modular Model Clusters

[0049] This step establishes and organizes the various sub-models required for digital twin simulation.

[0050] S101: Physical Process Decomposition and Modeling Requirements Definition. The target gas-steam combined cycle unit is decoupled into three main thermodynamic processes: the gas turbine thermodynamic cycle process, the waste heat boiler steam generation process, and the steam turbine expansion and work process.

[0051] Define at least two types of modeling requirements for each process: one type focuses on simulation accuracy and the integrity of physical mechanisms, used for steady-state fine analysis and economic optimization; the other type focuses on computational speed and dynamic response capability, used for rapid prediction and control optimization of transient processes.

[0052] S102: Specific development of multi-fidelity sub-models. To meet the above requirements, sub-models with different mathematical structures and complexities were developed. All models were implemented in a procedural or scripted manner and had clearly defined input and output interfaces.

[0053] Furthermore, for the gas turbine thermodynamic cycle process, a mechanism sub-model based on a set of nonlinear differential equations and a fast response sub-model based on real-time operating condition lookup tables or simplified formulas are established.

[0054] The inputs to the mechanistic sub-model include fuel flow commands. (Unit: kg / s), Ambient temperature (Unit: °C), Inlet guide vane opening (IGV) (Unit: %), Output including exhaust temperature (Unit: °C) Output power (Unit: MW), exhaust flow rate (Unit: kg / s);

[0055] The input and output physical quantities of the fast response sub-model are the same as those of the mechanism sub-model, but the internal calculation relationship is simplified, resulting in shorter calculation time.

[0056] Furthermore, for the steam generation process of waste heat boilers, a high-fidelity dynamic sub-model based on the multi-node distributed parameter method and a low-order dynamic sub-model based on the lumped parameter method are established.

[0057] The high-fidelity dynamic sub-model discretizes each heated surface into multiple micro-elements along the working fluid flow direction and solves the distributed parameter conservation equation.

[0058] The low-order dynamic sub-model simplifies the entire heat exchange zone into one or a few concentrated cavities.

[0059] Furthermore, for the steam turbine expansion process, a detailed dynamic sub-model considering the characteristics of the regulating stage and the volume effect, as well as a quasi-steady-state sub-model based on the fitting of the steam flow-power characteristic curve, are established.

[0060] S103: Model Standardization, Encapsulation, and Storage Management. Standardize and encapsulate all sub-models developed in step S102. Assign a unique identifier to each model instance and record its metadata according to a preset format. Metadata includes: model type, performance label ("High Fidelity" or "Fast"), list of input variables (name, unit, upper and lower limits), and list of output variables (name, unit).

[0061] After encapsulation, the model program file and its metadata are registered and stored in the system's model management module. This module provides an application programming interface for querying, retrieving, and loading based on model identifiers or attributes, in order to support the real-time scheduling and loading of corresponding sub-models according to the running mode instructions.

[0062] S200: Pre-configured optimization strategy and decision-making logic

[0063] This step configures the various parameters and logical rules required for the system to run.

[0064] S201: Setting State Feature Values ​​and Thresholds. In the configuration file of the state monitoring and decision-making module, set the algorithm used to calculate state feature values ​​and the corresponding threshold parameters.

[0065] Set load command change rate The calculation formula is ,in Target load (unit: MW). The sampling period (unit: s).

[0066] Set the main steam pressure tracking deviation change rate The calculation formula is: ,in , Actual value of main steam pressure (unit: MPa). Main steam pressure setpoint (unit: MPa).

[0067] At the same time, a first threshold is set. (Pe is the rated power), second threshold Third threshold Fourth threshold and steady-state confirmation time parameters .

[0068] S202: Configuration of optimization problem for steady-state optimization operation mode. In the optimization execution module, configure the optimization problem for steady-state optimization operation mode.

[0069] The optimization objective is a comprehensive function that includes fuel costs and environmental costs, and in this implementation, it is specifically defined as: ,in and These are the weighting coefficients. Total fuel flow predicted for the digital twin (unit: kg / s). The predicted total power (unit: MW). The predicted concentration of nitrogen oxides in exhaust gas (unit: ppm).

[0070] The constraints are configured as a series of inequalities, for example , , .

[0071] S203: Rolling optimization parameter configuration for transient tracking optimization mode. In the optimization execution module, configure model prediction control parameters for the transient tracking optimization mode, including the prediction time domain length. Controlling the length of the time domain Output error weight matrix Controlling the incremental weight matrix .

[0072] Simultaneously configure process constraints: control quantity amplitude constraints Control variable change rate constraint .

[0073] Phase Two: Online Dynamic Optimization of the Operational Closed Loop

[0074] This phase describes the continuous automated operation process of the system after it is connected to an actual generating unit.

[0075] S300: Real-time data synchronization and state characteristic calculation

[0076] The system exchanges data periodically with the unit's distributed control system through a data acquisition interface.

[0077] S301: Real-time data acquisition and preprocessing. At the beginning of each control cycle, the system reads the unit's current operating data, including actual power output. (Unit: MW) Actual value of main steam pressure (Unit: MPa), Gas turbine exhaust temperature (Unit: °C) Fuel valve position feedback (Unit: %) etc.

[0078] Simultaneously, external command data is read, including target load commands. (Unit: MW) and main steam pressure setpoint (Unit: MPa). The raw data is filtered and outlier removed.

[0079] S302: Online calculation of operating status characteristics. Based on the preprocessed data, the load command change rate at the current moment is calculated according to the formula preset in S201. and the rate of change of main steam pressure deviation .

[0080] S400: Operational Status Assessment and Model Reconstruction Decisions

[0081] The status monitoring and decision-making module judges the current and future situation based on preset logic and generates model reconstruction instructions.

[0082] S401: Pattern decision based on real-time feature quantities. This module has built-in decision logic. In each cycle, the following logic is executed:

[0083] like and The condition is maintained continuously for more than If so, the judgment will enter the steady-state optimization operation mode;

[0084] like or If any condition is met, the system will immediately enter the transient tracking optimization operation mode. The decision result will be output as a mode command signal.

[0085] Furthermore, when the decision is made to operate in steady-state optimization mode, a first switching instruction is generated. This instruction contains mapping information pointing to the high-fidelity sub-model identifiers required for each thermodynamic process.

[0086] Furthermore, when the decision is made to use the transient tracking optimization operation mode, a second switching instruction is generated. This instruction contains mapping information pointing to the fast sub-model identifiers required for each thermodynamic process.

[0087] S402: Pre-decision in conjunction with scheduling plan.

[0088] In a preferred implementation, when the condition monitoring and decision-making module generates the second switching instruction, it considers either the known future load plan of the unit or the current instruction. Based on short-term forecasts, if it is determined that the unit is about to enter the transient tracking optimization operation mode, a second switching instruction will be generated in advance.

[0089] In practice, future updates can be accessed through the plant-level monitoring information system. Calculate the average rate of change of the planned load curve over the most recent period (the next minute) using the 10-minute load plan curve. .

[0090] like Then, regardless of the current real-time feature values, a second switching instruction is immediately generated. Short-term prediction can also be achieved by first-order preservation of the current instruction or linear extrapolation.

[0091] S500: Dynamic Reconstruction and Real-Time Simulation of Digital Twins

[0092] Based on the first or second switching instruction generated by S400, the digital twin simulation module performs dynamic reconstruction of the digital twin.

[0093] S501: Instruction parsing and target model set determination. The digital twin simulation module parses the received mode instructions, and the module has a pre-stored running mode-sub-model mapping table.

[0094] For example, when the first switching instruction is received, the target sub-model identifier list is obtained by querying the mapping table: [Gas Turbine_High Fidelity, Waste Heat Boiler_High Fidelity, Steam Turbine_Detailed Dynamics].

[0095] When the second switching command is received, the following list is obtained: [Gas turbine_fast, waste heat boiler_low-order dynamic, steam turbine_quasi-steady state].

[0096] S502: Sub-model loading and coupling integration. The digital twin simulation module executes the following dynamic assembly process based on the received model switching instructions:

[0097] S502.1: Determine the set of target sub-models to be called based on the model switching instruction.

[0098] S502.2: Send a request to the model management module to load the corresponding sub-model program into the runtime memory and instantiate it based on the model identifier in the target set.

[0099] S502.3: Couple the input and output interfaces of each loaded sub-model according to the actual connection relationship of the thermodynamic process and the direction of material and energy transfer.

[0100] During implementation, data mapping relationships are established in memory through pointer or variable binding. When coupling connections are made, it is necessary to ensure that each sub-model satisfies the physical constraints of mass and energy conservation at the coupling interface.

[0101] Before simulation initialization, all coupled variables are checked for consistency, and the numerical solution stability of the overall model after combination is ensured by setting appropriate simulation step size and numerical integration algorithm.

[0102] S503: Simulation Initialization and Forward Prediction. The current real-time snapshot data of the unit obtained from S301 is used to initialize the state of the coupled overall digital twin model.

[0103] After initialization, the digital twin operates at a simulation step size. Running, predicting the future period of time The state evolution trajectory of the internal unit and the prediction results form a continuous data stream, which is provided to the optimization module.

[0104] S600: Optimization Strategy Matching and Online Solving

[0105] The optimization execution module receives the current operating mode instructions and the prediction data from the digital twin, and performs the corresponding optimization calculations.

[0106] S601: Optimization solution in steady-state optimization mode. When the input mode command is steady-state optimization mode, the steady-state optimizer is activated.

[0107] As a preferred implementation method, the steady-state optimization calculation uses a comprehensive function that includes fuel cost, power generation revenue and environmental cost as the optimization objective;

[0108] Based on the high-fidelity prediction data output by the digital twin simulation module, and under the condition of satisfying the safety operation constraints of all equipment in the unit, the steady-state operating point that optimizes the objective is solved, and the setting parameters corresponding to the steady-state operating point are output.

[0109] The specific implementation process is as follows: In each iteration, the steady-state optimizer generates a set of tentative setpoint vectors. This vector is then sent to the reconstructed high-fidelity digital twin. The digital twin, using the current real-time state as initial conditions, simulates to a new steady state and returns the predicted values ​​and constraint violations in that steady state.

[0110] The optimizer calculates the objective function value based on the returned information. Based on the degree of constraint violation, an optimization algorithm is used to adjust and generate the next set of trial values. This iterative process continues until a value is found that satisfies all constraints and makes... Minimize the optimal setting value This optimization process is performed periodically.

[0111] S602: Rolling optimization solution in transient tracking optimization mode. When the input mode command is transient tracking optimization mode, the transient optimizer is activated.

[0112] As a preferred implementation, the transient rolling optimization calculation at the beginning of each control cycle takes minimizing the prediction tracking error of the key controlled parameters within a finite future time period as the optimization objective.

[0113] Based on the rapid prediction data output by the digital twin simulation module, the optimal control sequence is solved in a rolling manner under the condition of satisfying the dynamic process variable change rate constraint, and the control command to be executed in the sequence is output.

[0114] The specific implementation process is as follows: In each control cycle The transient optimizer uses the current measured state of the unit as its initial condition, and it will include a future... A trial sequence of control quantities Send to the reconstructed rapid digital twin.

[0115] Twins integrate forward over simulation time Predict the corresponding output trajectory within seconds. The optimizer calculates the predicted trajectory. With the target trajectory The objective function value is calculated using the weighted sum of squared errors, and then adjusted using an optimization algorithm to finally obtain the optimal control sequence that satisfies all process constraints. .

[0116] The optimizer will use the first control variable in the sequence This will be the output for this cycle.

[0117] S700: Optimize the security verification and execution of instructions

[0118] This step completes the transmission of control commands from the virtual space to the physical device, achieving closed-loop control.

[0119] S701: Post-processing and safety limiting of output instructions. Performs post-processing on optimized output instructions generated by S601 or S602.

[0120] The processing includes: limiting the output of the setpoint to ensure it is within the acceptable range of the actuator; and limiting the rate of change of the directly controlled quantity.

[0121] Finally, the processed instructions are compared with the hard protection settings of the unit's underlying safety monitoring system. This step constitutes a safety check to ensure that the optimized instructions are always within the safety threshold of the unit's underlying protection system.

[0122] S702: Command Issuance and New Loop. The optimized command, validated through final safety verification, is issued to the unit's basic distributed control system via industrial communication protocol. The basic distributed control system drives the actuators in the field, such as fuel regulating valves and turbine inlet steam regulating valves. The system then waits for the next data sampling cycle and begins a new closed-loop process from step S300.

[0123] The specific embodiments described above demonstrate a feasible implementation of the method and system. Any equivalent transformations or substitutions that can be conceived by those skilled in the art based on the logic and principles disclosed in this solution without creative effort should be included within the scope of the protection intent of this solution.

[0124] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other.

[0125] 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 dynamic optimization operation system for a gas-steam combined cycle unit based on digital twins, characterized in that, include: The model management module is used to store simulation sub-models for gas turbine, waste heat boiler and steam turbine processes, where each process stores at least one high-fidelity sub-model and one fast sub-model; The status monitoring and decision-making module is used to collect unit operating data and external commands, calculate the load command change rate and the key parameter deviation change rate, compare the change rate with a preset threshold, and generate a first switching command or a second switching command based on the comparison result. The digital twin simulation module is used to call and couple the corresponding high-fidelity sub-model according to the first switching instruction, or to call and couple the corresponding fast sub-model according to the second switching instruction, so as to form a digital twin under the current working condition and perform simulation prediction. The optimization execution module is used to perform steady-state optimization based on the prediction data of the high-fidelity sub-model with economic indicators as the objective when the first switching instruction is generated; and to perform rolling optimization based on the prediction data of the fast sub-model with dynamic tracking performance as the objective when the second switching instruction is generated.

2. The dynamic optimization operation system for gas-steam combined cycle units based on digital twins according to claim 1, characterized in that, The specific method by which the status monitoring and decision-making module generates instructions based on the comparison results is as follows: S1: Calculate the rate of change of the load command and the rate of change of the deviation between the actual value and the set value of the main steam pressure at the current moment; S2: If the load command change rate continues to be lower than the first threshold and the deviation change rate continues to be lower than the second threshold, then a first switching command is generated; S3: If the load command change rate exceeds the third threshold or the deviation change rate exceeds the fourth threshold, a second switching command is generated.

3. The dynamic optimization operation system for gas-steam combined cycle units based on digital twins according to claim 2, characterized in that, When the status monitoring and decision-making module executes step S3, if it determines, based on the known future load plan or the prediction based on the current instruction, that the load instruction change rate is about to exceed the third threshold, it generates a second switching instruction in advance.

4. The dynamic optimization operation system for gas-steam combined cycle units based on digital twins according to claim 1, characterized in that, The specific process by which the digital twin simulation module calls and couples sub-models according to the switching command is as follows: S1: Parse the switching instruction and determine the set of high-fidelity or fast sub-model identifiers to be called; S2: Load the sub-model corresponding to the identifier from the model management module into the runtime memory; S3: Connect the input and output interfaces of the loaded sub-model according to the actual energy flow and working fluid flow direction of the unit.

5. The dynamic optimization operation system for gas-steam combined cycle units based on digital twins according to claim 1, characterized in that, The specific method by which the optimization execution module performs steady-state optimization is as follows: S1: The optimization objective is a comprehensive cost function that includes fuel consumption and emission indicators; S2: Based on the prediction data of the high-fidelity digital twin, the optimal steady-state setpoint is obtained by solving the problem under the condition of satisfying the safety operation boundary of the equipment.

6. The dynamic optimization operation system for gas-steam combined cycle units based on digital twins according to claim 1, characterized in that, The specific method by which the optimization execution module performs rolling optimization is as follows: S1: The optimization objective is to minimize the prediction and tracking error of key parameters within a finite future time period. S2: Based on the prediction data of the fast digital twin, a rolling solution is performed in each control cycle, and the control command for the next cycle is output.

7. A method for dynamic optimization operation of a gas-steam combined cycle unit based on digital twins, characterized in that, Including the following steps: S100: Establish and store high-fidelity and fast sub-models of gas turbine, waste heat boiler and steam turbine processes; S200: Rate of change of load commands and rate of change of deviation of key parameters for real-time computer groups; S300: Compare the rate of change with a preset threshold to generate a first switching instruction or a second switching instruction; S400: Based on the generated switching instructions, dynamically select and couple the corresponding high-fidelity or fast sub-model to form a digital twin for simulation; S500: If a first switching instruction is generated, steady-state optimization is performed based on the high-fidelity simulation results; if a second switching instruction is generated, rolling optimization is performed based on the fast simulation results.

8. The method for dynamic optimization operation of gas-steam combined cycle units based on digital twins according to claim 7, characterized in that, The conditions for generating the first switching instruction in step S300 are: the load instruction change rate is continuously lower than the first threshold and the key parameter deviation change rate is continuously lower than the second threshold; the conditions for generating the second switching instruction are: the load instruction change rate exceeds the third threshold or the key parameter deviation change rate exceeds the fourth threshold.

9. The method for dynamic optimization operation of gas-steam combined cycle units based on digital twins according to claim 7, characterized in that, The steady-state optimization described in step S500 specifically involves solving for the steady-state setpoint with the overall operating cost as the objective; the rolling optimization specifically involves rolling the solution for the recent control commands with the dynamic tracking performance as the objective.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the method as described in any one of claims 7 to 9.

Images