CN-122021280-A - Simplified dynamic modeling method for heat storage tank based on fusion of model driving and data driving

Abstract

The invention provides a simplified dynamic modeling method of a heat storage tank based on fusion of model driving and data driving, which comprises the steps of constructing a first principle model based on structural parameters and operation constraint of the heat storage tank, constructing a linear state space model based on the first principle model through a system identification method, generating variable working condition input instruction data of the heat storage tank by adopting a pseudo-random binary signal as input of the first principle model and the state space model of the heat storage tank, sampling to obtain an input and output parameter simulation data set of the heat storage tank, combining actual operation data of the heat storage tank to jointly form a data driving model training data set, constructing and training to obtain an artificial neural network data driving model based on the operation parameter training data set, and serially connecting the state space model and an ANN data driving model to construct a hybrid driving model frame. The invention lightens the model structure of the heat storage tank, improves the modeling precision and provides a model foundation for the research of a heat storage tank control method.

Inventors

• XU BIN
• GUO ZHEN
• TANG LINGLING
• WANG NING
• ZHAO WANG
• SUN JIAPIN
• LI DONG
• YU JIABIN
• WU QIONG
• YIN XINYU
• ZHANG JUNLI
• LI YIGUO
• CHEN HUI
• YU ZUOXING

Assignees

• 华能中盐(常州)储能有限公司
• 华能国际电力江苏能源开发有限公司
• 东南大学

Dates

Publication Date

20260512

Application Date

20260121

Claims (10)

1. 1. The simplified dynamic modeling method for the heat storage tank based on fusion of model driving and data driving is characterized by comprising the following steps of: acquiring a physical structure and operation constraint of a heat storage tank, and constructing a first principle model of the heat storage tank as an operation reference model of the heat storage tank; a state space model is obtained through system identification based on the first sexual principle model and is used for describing the dynamic characteristics of the heat storage tank; based on the first sexual principle model and the state space model, constructing an operation parameter data set required by data driving model training by combining actual operation data; constructing an artificial neural network ANN model based on the operation parameter data set, and correcting nonlinear errors between the state space model and the first sexual principle model; and (3) connecting the state space model and the ANN model in series to form a model and data hybrid driving modeling framework, and carrying out modeling simulation analysis on the heat storage tank.
2. 2. The method of claim 1, wherein the obtaining the physical structure and the operation constraint of the thermal storage tank, constructing a first principle model of the thermal storage tank as an operation reference model of the thermal storage tank, comprises: The mathematical expression of the first principle model for constructing the heat storage tank based on mass conservation and energy conservation is as follows: Wherein, the The mass of the heat storage water in the heat storage tank is kg; And The flow rate of the hot water flowing into and out of the heat storage tank is kg/s respectively; the temperature of hot water in the heat storage tank is DEG C; for the temperature of the hot water flowing into the thermal storage tank, °c; the specific heat capacity of hot water in the heat storage tank is kJ/(kg DEG C); The heat of the heat storage tank to the environment is dissipated, and kW; the heat dissipation coefficient is kW/(m2·DEG C); Is the heat dissipation area, m2; is ambient temperature, °c; The first sexual principle model is used as an operation reference model.
3. 3. The method of claim 2, wherein establishing a state space model expression for the thermal storage tank is: Wherein the method comprises the steps of 、 、 、 The system matrix, the input matrix, the output matrix and the transfer matrix are respectively; the input parameters of the heat storage tank are the flow rate and the temperature of hot water flowing into the heat storage tank and the flow rate of hot water flowing out of the heat storage tank; Is a state variable matrix; and for outputting the parameter matrix, the output parameter of the heat storage tank is the hot water temperature of the heat storage tank, and the ratio of the current water volume to the total volume of the heat storage tank, namely the SOC value.
4. 4. A method according to claim 3, wherein constructing the set of operational parameter data required for data-driven model training in combination with actual operational data based on the first principles model and the state space model comprises: The method comprises the steps of adopting model simulation data as a training data set, adopting a pseudo-random binary signal with random repetition duration to generate input parameter data of the heat storage tank, using the input parameter data as input to a first sexual principle model and a state space model of the heat storage tank to obtain corresponding output parameter data, and finally obtaining a parameter set of input parameters and output parameters through sampling.
5. 5. The method of claim 4, wherein the artificial neural network ANN model constructed based on the operating parameter data set is expressed as: Wherein, the 、 、 The weight matrix is the weight matrix of the neurons of the input layer, the k hidden layer and the output layer; 、 、 bias vectors for neurons of an input layer, a k hidden layer and an output layer; is a nonlinear activation function; Inputting a feature data set for training; hiding layer neuron set vectors for a k-th layer; collecting vectors for neurons of a last layer of hidden layers; Is an output feature.
6. 6. The method of claim 5, wherein concatenating the state space model with the ANN model forms a model and data hybrid drive modeling framework comprising: the ANN model takes output parameters of the state space model and input parameters of the state space model as model inputs, corrects residual errors of the state space model and the first sexual principle model, and adds the output of the ANN model as a correction term and an output result of the state space model to be used as a final output of the whole hybrid driving model, wherein the output expression of the hybrid model is as follows: Wherein, the 、 And The final output vector of the hybrid driving model, the output vector of the state space model and the output vector of the ANN model are respectively; 、 Input vectors of the state space model and the ANN model are respectively; the parameter set is a parameter set of a state space model and is obtained by data identification; the model is a parameter set of the ANN model and is obtained by training a state space model output and a residual error data set output based on a first sex principle dynamic model.
7. 7. The simplified dynamic modeling device for the heat storage tank based on fusion of model driving and data driving is characterized by comprising the following components: the first module is used for acquiring the physical structure and the operation constraint of the heat storage tank, constructing a first sex principle model of the heat storage tank and taking the first sex principle model as an operation reference model of the heat storage tank; The second module is used for obtaining a state space model based on the first sexual principle model through system identification and describing the dynamic characteristics of the heat storage tank; the third module is used for constructing an operation parameter data set required by data driving model training by combining actual operation data based on the first sexual principle model and the state space model; The fourth module is used for constructing an artificial neural network ANN model based on the operation parameter data set and correcting nonlinear errors between the state space model and the first principle model; and the fifth module is used for connecting the state space model and the ANN model in series to form a model and data hybrid driving modeling framework, and carrying out modeling simulation analysis on the heat storage tank.
8. 8. An electronic device comprising a processor and a memory communicatively coupled to the processor; The memory stores computer-executable instructions; the processor executes computer-executable instructions stored in the memory to implement the method of any one of claims 1-6.
9. 9. A computer readable storage medium having stored therein computer executable instructions which when executed by a processor are adapted to carry out the method of any one of claims 1-6.
10. 10. A computer program product comprising a computer program which, when executed by a processor, implements the method of any of claims 1-6.

Description

Simplified dynamic modeling method for heat storage tank based on fusion of model driving and data driving Technical Field The invention relates to the technical field of dynamic modeling of energy equipment, in particular to a simplified dynamic modeling method of a heat storage tank based on fusion of model driving and data driving. Background The heat storage tank is used as core equipment of a heat energy storage system and plays a key role in the fields of solar thermal power generation, industrial waste heat recovery, compressed Air Energy Storage (CAES) and the like. The dynamic characteristics directly affect the energy management efficiency, the running stability and the economy of the system. At present, modeling methods of a heat storage tank mainly comprise a pure mechanism modeling method and a pure data driving modeling method. The pure mechanism modeling method is mainly based on the physical conservation law, and can accurately describe the fluid dynamics and the heat transfer process in the tank, but the calculation complexity is high, for example, the single dynamic simulation of the multi-node CFD model of the heat storage tank consumes several hours, so that the timeliness requirement of real-time control or daily scheduling optimization is difficult to meet. Meanwhile, the precision of the pure mechanism model depends on accurate boundary conditions, and the parameters are difficult to directly measure in engineering practice, so that the practicability of the model is reduced. The pure data driving model only depends on the heat storage tank operation data to fit the operation characteristics, the modeling method is simple, but high-quality training data covering all working conditions is needed, and the model lacks clear mechanism explanation. To balance computational efficiency and accuracy, simplified mechanism models are often used in engineering to replace complex mechanism models, but they result in loss of key dynamics and reduced accuracy. In view of the above, it is difficult to combine the model accuracy and the calculation efficiency in the existing thermal storage tank dynamic modeling method, and a new modeling method is needed to construct a lightweight and high-accuracy thermal storage tank model. Disclosure of Invention The present invention aims to solve at least one of the technical problems in the related art to some extent. Therefore, a first object of the present invention is to provide a simplified dynamic modeling method for a thermal storage tank based on fusion of model driving and data driving, wherein a state space model is used for describing dynamic characteristics of the thermal storage tank, and a data driving model is used for compensating and modeling nonlinear errors of simulation of the state space model, so as to construct a high-precision lightweight thermal storage tank dynamic model. The second aim of the invention is to provide a simplified dynamic modeling device of the heat storage tank based on fusion of model driving and data driving. A third object of the present invention is to propose an electronic device. A fourth object of the present invention is to propose a computer readable storage medium. A fifth object of the invention is to propose a computer programme product. To achieve the above objective, an embodiment of a first aspect of the present invention provides a simplified dynamic modeling method for a thermal storage tank based on fusion of model driving and data driving, including: acquiring a physical structure and operation constraint of a heat storage tank, and constructing a first principle model of the heat storage tank as an operation reference model of the heat storage tank; a state space model is obtained through system identification based on the first sexual principle model and is used for describing the dynamic characteristics of the heat storage tank; based on the first sexual principle model and the state space model, constructing an operation parameter data set required by data driving model training by combining actual operation data; constructing an artificial neural network ANN model based on the operation parameter data set, and correcting nonlinear errors between the state space model and the first sexual principle model; and (3) connecting the state space model and the ANN model in series to form a model and data hybrid driving modeling framework, and carrying out modeling simulation analysis on the heat storage tank. Optionally, the obtaining the physical structure and the operation constraint of the heat storage tank, and constructing a first sexual principle model of the heat storage tank as an operation reference model of the heat storage tank, includes: The mathematical expression of the first principle model for constructing the heat storage tank based on mass conservation and energy conservation is as follows: Wherein, the The mass of the heat storage water in the heat storage tank is kg; And The flow ra