CN-122020013-A - Digital twin model construction method and device for coal-fired generator set

CN122020013ACN 122020013 ACN122020013 ACN 122020013ACN-122020013-A

Abstract

The application discloses a digital twin model construction method and device for a coal-fired generator set. The method comprises the steps of constructing a mechanism model for describing the operation state of the coal-fired power unit based on the structure and the operation mechanism of the coal-fired power unit, predicting historical input data based on the mechanism model to obtain mechanism prediction output corresponding to each piece of historical input data, performing model training on an initial data driving model based on actual measurement output and mechanism prediction output of the same piece of historical input data and a characteristic vector of unit operation to obtain a target data driving model, and constructing a target digital twin model based on the mechanism model and the target data driving model. The method can accurately and reasonably construct the digital twin model of the coal-fired power generator unit, and lays a foundation for the follow-up accurate prediction of the running state of the coal-fired power generator unit.

Inventors

BAO ZHEN
ZHANG ZHIGANG
NIU HAIMING
ZHONG ZHEN
FANG FANG
YAN GUANGTAO
ZHANG DONGMING
JIANG HEQING

Assignees

国能智深控制技术有限公司

Dates

Publication Date: 20260512
Application Date: 20251205

Claims (10)

1. The digital twin model construction method of the coal-fired generator set is characterized by comprising the following steps of: Based on the structure and the operation mechanism of the coal-fired power generation unit, constructing a mechanism model for describing the operation state of the coal-fired power generation unit; Predicting the historical input data based on the mechanism model to obtain mechanism prediction output corresponding to each historical input data; Model training is carried out on the initial data driving model based on the actual measurement output and the mechanism prediction output of the same historical input data and the characteristic vector of unit operation, and a target data driving model is obtained; And constructing a target digital twin model based on the mechanism model and the target data driving model.
2. The method of claim 1, wherein the building a mechanism model for describing the operation state of the coal-fired unit based on the thermodynamic system structure and the operation mechanism of the coal-fired unit specifically comprises: aiming at the thermodynamic system structure and the operation mechanism of the coal-fired unit, acquiring the state vector of the coal-fired unit; Based on the state vector and the control input vector of the coal-fired unit, establishing a mechanism model for describing the operation state of the coal-fired unit; The state vector comprises any several energy storage variables of the temperature, pressure, drum water quantity and turbine rotating speed of each section of the boiler; The control input vector comprises any one of a fuel quantity, a water supply quantity, an air quantity and a valve opening degree.
3. The method of claim 1, wherein the model training the initial data-driven model based on the measured output and the mechanism prediction output of the same historical input data, and the feature vector of the unit operation, to obtain the target data-driven model, specifically comprises: constructing a residual error sample based on the actual measurement output and the mechanism prediction output of the same historical input data; and performing model training on the initial data driving model based on the residual error sample and the characteristic vector of unit operation to obtain a target data driving model.
4. The method of claim 1, wherein after obtaining the target digital twin model, the method further comprises filtering the prediction data output by the target digital twin model based on any one of unscented Kalman filtering, extended Kalman filtering, ensemble Kalman filtering, and particle filtering.
5. The method of claim 4, wherein the filtering the predicted data output by the target digital twin model based on unscented kalman filtering specifically comprises: Constructing an augmented state vector based on the state vector of the mechanism model and mechanism model parameters; aiming at the current moment, the augmented state vector of the previous moment is input into the target digital twin model to obtain the current model output; Acquiring an actual measured value of a coal-fired generator set at the current moment; And updating the augmented state vector and the filtering parameters based on the actual measured value and the current model output so as to carry out filtering processing on the predicted data output by the target digital twin model based on the updated filtering parameters.
6. The method of claim 3, wherein after constructing the target digital twin model, the method further comprises: acquiring the current actual output of the coal-fired generator set and the current mechanism prediction output of a mechanism model in the target digital twin model; updating the residual error sample based on the current actual output and the current mechanism prediction output, so as to perform incremental training on the target data driving model based on the updated residual error sample, and obtain an updated target data driving model; and optimizing the target digital twin model based on the updated target data driving model to obtain an optimized target digital twin model.
7. The method of claim 1, wherein after obtaining the target digital twin model, the method further comprises: Determining a current model parameter predicted value of a current time period based on current actual model parameters of a mechanism model in the target digital twin model in the current time period; and determining whether the mechanism model meets the preset parameter optimization conditions or not based on the current model parameter predicted value and a preset model parameter standard value, and adjusting model parameters of the mechanism model when the mechanism model meets the parameter optimization conditions.
8. The digital twin model construction device of the coal-fired generator set is characterized by comprising: the building module is used for building a mechanism model for describing the operation state of the coal-fired power generation unit based on the structure and the operation mechanism of the coal-fired power generation unit; the prediction module is used for predicting the historical input data based on the mechanism model to obtain mechanism prediction output corresponding to each historical input data; The data driving model training module is used for carrying out model training on the initial data driving model based on the actual measurement output and the mechanism prediction output of the same historical input data and the characteristic vector of the unit operation to obtain a target data driving model; And the construction module is used for constructing a target digital twin model based on the mechanism model and the target data driving model.
9. A storage medium storing a computer program which when executed by a processor performs the steps of the method of constructing a digital twin model of a coal-fired power generation unit according to any of claims 1-7.
10. An electronic device comprising at least a memory, a processor, the memory having stored thereon a computer program, the processor, when executing the computer program on the memory, implementing the steps of the method for constructing a digital twin model of a coal-fired power generation unit according to any of the preceding claims 1-7.

Description

Digital twin model construction method and device for coal-fired generator set Technical Field The invention relates to the field of modeling of thermal generator sets, in particular to a method and a device for constructing a digital twin model of a coal-fired generator set. Background The coal-fired generator set is used as a typical complex thermodynamic system, the operation process of the coal-fired generator set is coupled with a plurality of subsystems such as boiler combustion, steam-water circulation, steam turbine work, power generation, electric control system and the like, and the coal-fired generator set relates to multiple subjects such as thermodynamics, control theory and the like, and has obvious interaction of multiple physical fields and multiple time scales. In the running process of the coal-fired unit, the coal-fired unit faces complex working conditions such as frequent start and stop, deep peak shaving and the like, and higher requirements are put forward on modeling precision, instantaneity and self-adaption capability of the unit. Under the background, the digital twin technology is used for constructing a coal-fired unit model which can cover all working conditions and is updated in a self-adaptive mode, so that the safety, economy and operation flexibility of the unit can be obviously improved, and basic support is provided for operation optimization, fault prediction and intelligent operation and maintenance. However, existing coal-fired unit models typically employ white-box modeling methods based on physical mechanisms and black-box modeling methods based on data drives. A white-box modeling method based on a physical mechanism relies on basic physical laws such as thermodynamics and hydrodynamics to establish a unit dynamics model comprising a boiler, a steam turbine, a generator and a control system. The model has the advantages of strong physical interpretability, clear structure and the like, but a large number of simplifying assumptions are needed for the system, and parameter identification is carried out depending on test data. When the unit operation conditions are obviously changed (such as deep peak shaving and quick start-stop) or the equipment is worn and aged, the original parameters are difficult to update in time, so that the model accuracy is obviously reduced. A black box modeling method based on data driving utilizes historical operation data to directly fit nonlinear mapping between input and output through a system identification or machine learning method (such as a neural network, a gradient lifting tree and the like). The model has stronger function approximation capability, and is easy to obtain higher fitting precision in complex nonlinear scenes. However, due to lack of physical constraints, its extrapolation capability is poor, reliability under conditions other than training data is difficult to guarantee, and model interpretability is weak. Therefore, a digital twin model construction method of a coal-fired power generator unit is needed urgently to solve the problems that in the prior art, the construction of a coal-fired power generator unit model is unreasonable and inaccurate, and the operation state of the coal-fired power generator unit cannot be accurately predicted. Disclosure of Invention In view of the above, the invention provides a digital twin model construction method of a coal-fired generator set, which mainly aims to solve the problems that the current coal-fired generator set model construction is not reasonable and accurate enough and the operation state of the coal-fired generator set cannot be accurately predicted. In order to solve the problems, the invention provides a digital twin model construction method of a coal-fired generator set, which comprises the following steps: Based on the structure and the operation mechanism of the coal-fired power generation unit, constructing a mechanism model for describing the operation state of the coal-fired power generation unit; Predicting the historical input data based on the mechanism model to obtain mechanism prediction output corresponding to each historical input data; Model training is carried out on the initial data driving model based on the actual measurement output and the mechanism prediction output of the same historical input data and the characteristic vector of unit operation, and a target data driving model is obtained; And constructing a target digital twin model based on the mechanism model and the target data driving model. Optionally, the establishment of the mechanism model for describing the operation state of the coal-fired unit based on the thermodynamic system structure and the operation mechanism of the coal-fired unit specifically comprises the following steps: aiming at the thermodynamic system structure and the operation mechanism of the coal-fired unit, acquiring the state vector of the coal-fired unit; Based on the state vector and the control input vecto