Dynamic system hybrid modeling method and device, electronic equipment and storage medium

Abstract

The application relates to the technical field of simulation modeling, and provides a dynamic system hybrid modeling method and device, electronic equipment and a storage medium. The method comprises the following steps: firstly, a data-driven model and a mechanism model of a target component in a dynamic system are established; then, input data is configured for an input variable of the target component, a first measurement value of an output variable of the target component is determined, the input data is input into the mechanism model, and a first output value determined by the mechanism model is obtained; thereafter, an error between the first output value and the first measurement value is calculated, a second measurement value of a measurable parameter of the target component and the error are input into the data-driven model, and a second output value of the data-driven model is obtained; finally, the first output value is compensated for error based on the second output value. The method can reduce the dependence on the mechanism model by compensating for error of the output of the mechanism model through the data-driven model, can obtain an accurate output of the mechanism model, and can further improve the accuracy and reliability of overall modeling.

Description

Technical Field

[0001] This invention relates to the field of simulation modeling technology, and in particular to a method, apparatus, electronic device, and storage medium for hybrid modeling of dynamic systems. Background Technology

[0002] Dynamic system modeling is the foundation for analyzing and controlling dynamic systems.

[0003] Currently, dynamic system modeling methods mainly include serial hybrid-driven modeling methods and parallel hybrid-driven modeling methods. Serial hybrid-driven modeling methods combine data-driven and mechanism-driven modeling methods in a series, meaning one method is used to build the model, and then another method is used to modify or optimize it. However, the hybrid models obtained by this method have poor accuracy.

[0004] Parallel hybrid-driven modeling methods combine data-driven and mechanism-driven modeling approaches in parallel. This involves simultaneously building models using both methods and then using a third method to fuse or select the best model from the two. The resulting hybrid model has a relatively simple structure and offers greater flexibility in modeling. However, the extrapolation and generalization performance of this hybrid model heavily relies on the mechanistic model. When the accuracy of the mechanistic model is poor, the model may deviate significantly from the actual research object. Summary of the Invention

[0005] This invention provides a method, apparatus, electronic device, and storage medium for dynamic system hybrid modeling, in order to overcome the deficiencies existing in the prior art.

[0006] This invention provides a method for modeling hybrid dynamic systems, comprising:

[0007] Establish data-driven and mechanistic models of target components in dynamic systems;

[0008] Input data is configured for the input variables of the target component, a first measured value of the output variables of the target component is determined, and the input data is input into the mechanism model to obtain the first output value determined by the mechanism model;

[0009] Calculate the error between the first output value and the first measured value, and input the second measured value of the measurable parameter of the target component and the error into the data-driven model to obtain the second output value of the data-driven model;

[0010] Based on the second output value, error compensation is applied to the first output value to obtain the third output value.

[0011] According to a dynamic system hybrid modeling method provided by the present invention, the mechanistic model and the data-driven model are connected in parallel and then in series with the mechanistic model; correspondingly, the step of performing error compensation on the first output value based on the second output value to obtain a third output value includes:

[0012] The mechanism model is optimized based on the third output value.

[0013] According to a dynamic system hybrid modeling method provided by the present invention, the data-driven model and the mechanism model are connected in series and then connected in parallel with the data-driven model; correspondingly, the mechanism model is determined based on the following steps:

[0014] Based on the data-driven model, the mechanism model is established.

[0015] According to the present invention, a method for modeling hybrid dynamic systems, wherein establishing the mechanism model based on the data-driven model includes:

[0016] The data-driven model is used as the mechanism model;

[0017] or,

[0018] A basic mechanism model is determined, and the third measured value of the measurable parameter of the target component is input into the data-driven model to obtain the fourth output value of the data-driven model;

[0019] Based on the third measurement value and the fourth output value, the values ​​of the mechanism parameters of the basic mechanism model are determined, and the mechanism model is constructed based on the values ​​of the mechanism parameters.

[0020] According to the dynamic system hybrid modeling method provided by the present invention, both the first measurement value and the second measurement value are obtained based on one-dimensional mechanism simulation.

[0021] According to the present invention, a method for modeling a dynamic system is provided, wherein the dynamic system is a power system and the target component is one of a condenser, a heat exchanger, a pump, and a valve.

[0022] According to the present invention, a dynamic system hybrid modeling method is provided, wherein the data-driven model is constructed based on a feedforward neural network or a long short-term memory neural network.

[0023] The present invention also provides a dynamic system hybrid modeling device, comprising:

[0024] The model building module is used to build data-driven models and mechanism models of target components in dynamic systems;

[0025] The data preparation module is used to configure input data for the input variables of the target component, determine the first measured value of the output variables of the target component, and input the input data into the mechanism model to obtain the first output value determined by the mechanism model.

[0026] An error calculation module is used to calculate the error between the first output value and the first measured value, and input the second measured value of the measurable parameter of the target component and the error into the data-driven model to obtain the second output value of the data-driven model;

[0027] An error compensation module is used to perform error compensation on the first output value based on the second output value to obtain a third output value.

[0028] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the dynamic system hybrid modeling method as described above.

[0029] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the dynamic system hybrid modeling method as described above.

[0030] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the dynamic system hybrid modeling method as described above.

[0031] This invention provides a method, apparatus, electronic device, and storage medium for hybrid modeling of dynamic systems. The method first establishes a data-driven model and a mechanistic model of the target component in the dynamic system. Then, it configures input data for the input variables of the target component, determines the first measured value of the output variables of the target component, and inputs the input data into the mechanistic model to obtain the first output value determined by the mechanistic model. Next, it calculates the error between the first output value and the first measured value, and inputs the second measured value of the measurable parameters of the target component and the error into the data-driven model to obtain the second output value of the data-driven model. Finally, based on the second output value, it performs error compensation on the first output value to obtain a third output value. This method reduces the dependence on the mechanistic model by using the data-driven model to compensate for errors in the output of the mechanistic model. Even when the mechanistic model is not clear, needs simplification, or is unknown, it can still obtain an accurate output of the mechanistic model, thereby improving the overall accuracy and reliability of the modeling. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on the drawings described below without creative effort.

[0033] Figure 1 This is a flowchart illustrating the dynamic system hybrid modeling method provided by the present invention;

[0034] Figure 2 This is one of the structural schematic diagrams of the power system provided by the present invention;

[0035] Figure 3 This is the second schematic diagram of the power system provided by the present invention;

[0036] Figure 4 This is a schematic diagram of the parallel-serial hybrid modeling structure in the dynamic system hybrid modeling method provided by the present invention;

[0037] Figure 5 This is a schematic diagram of the serial-parallel hybrid modeling structure in the dynamic system hybrid modeling method provided by the present invention;

[0038] Figure 6 This is a schematic diagram of the structure of the dynamic system hybrid modeling device provided by the present invention;

[0039] Figure 7 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0041] The terms "first" and "second" in the specification and claims of this invention may explicitly or implicitly include one or more of those features. In the description of the invention, unless otherwise stated, "a plurality of" means two or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0042] Because existing dynamic system modeling methods often produce hybrid models with poor accuracy or excessive reliance on mechanistic models, the resulting models suffer from low reliability and accuracy, hindering stable application. Therefore, this invention provides a hybrid modeling method for dynamic systems.

[0043] Figure 1 This is a flowchart illustrating a dynamic system hybrid modeling method provided in an embodiment of the present invention, such as... Figure 1 As shown, the method includes:

[0044] S1, Establish the data-driven model and mechanism model of the target component in the dynamic system;

[0045] S2, configure input data for the input variables of the target component, determine the first measured value of the output variables of the target component, and input the input data into the mechanism model to obtain the first output value determined by the mechanism model;

[0046] S3, calculate the error between the first output value and the first measured value, and input the second measured value of the measurable parameter of the target component and the error into the data-driven model to obtain the second output value of the data-driven model;

[0047] S4. Based on the second output value, the first output value is compensated for error to obtain the third output value.

[0048] Specifically, the dynamic system hybrid modeling method provided in this embodiment of the invention is executed by a dynamic system hybrid modeling device, which can be configured in a computer. The computer can be a local computer or a cloud computer. The local computer can be a computer, tablet, etc., and no specific limitation is made here.

[0049] First, perform step S1 to establish the data-driven model and mechanism model of the target component in the dynamic system. This dynamic system can be a power system.

[0050] like Figure 2 As shown, the power system may include a steam generator, a steam turbine, a generator, a condenser, a condensate pump, a feedwater pump, a water tank, and other equipment. The feedwater pump's output power absorbs heat transferred from the coolant in the steam generator and evaporates into steam. The qualified steam first enters the steam turbine to perform work. The exhaust steam, after completing its work, is discharged into the condenser and condenses into water. The condensate pump then sends the condensate to the water tank, and finally, the feedwater pump pumps it back into the steam generator, forming a closed-loop power cycle.

[0051] like Figure 3As shown, the power system may also include equipment such as a steam generator, a condenser, and a heat exchanger. Steam enters the condenser and condenses by contacting the liquid phase. After condensation, the condensate is distributed, with a portion entering the steam generator and the other portion entering the heat exchanger for further cooling. The cooled fluid portion returns to the condenser to cool new steam. The condenser in the power system is a mixing condenser, and the heat exchanger can be a shell-and-tube heat exchanger or a tube-and-tube heat exchanger. The heat exchanger is of the liquid-liquid heat exchange type, meaning that both the cold and hot sides of the heat exchanger are liquids.

[0052] The target component in the power system can be one of the following: condenser, heat exchanger, pump, and valve. The data-driven model of the target component can be constructed based on a feedforward (BP) neural network or a long short-term memory (LSTM) neural network.

[0053] For example, a data-driven model for a pump can use the speed-time curve in the test curve diagram during pump startup as modeling data, extract it, and input it into a BP neural network for training. Alternatively, it can use the flow-head curve, flow-efficiency curve, and flow-power curve in the external characteristic prediction diagram as modeling data, extract it, and input it into an LSTM neural network for training.

[0054] The data-driven model of a valve can be obtained by extracting the pressure-time curve before the valve from the average pressure change curve before the valve and inputting it into a BP neural network model for training, or by extracting the pressure-time curve after the valve from the average pressure change curve after the valve and inputting it into an LSTM neural network model for training.

[0055] The condenser can be a mixing condenser, which mainly relies on spraying cooling water to form a water film, which directly contacts the turbine exhaust steam for heat exchange. First, the steam-air mixture passes through a water chamber, i.e., a steam cooling channel, where the exhaust steam is cooled into water. After passing through the cooling channel, a portion of the steam remains uncooled. To reduce heat loss and condensate loss, an ejector draws this uncondensed steam-air mixture into the aftercooling zone for further cooling. A small portion of the cooling water enters the aftercooler through the water chamber, where a water film formed by a water spray plate further contacts the mixed gas, causing the remaining steam to cool and condense. Finally, the air is extracted through the exhaust port and discharged into the atmosphere via an ejector. The amount of cooling water entering the aftercooler is only a small portion, accounting for 5% to 10% of the total cooling water volume.

[0056] When constructing the mechanistic model of the condenser, let the mass flow rate be W. L The temperature is t WCooling water enters the condenser, and the exhaust volume is D. k Steam heated to saturation temperature t s The steam is simultaneously condensed into water, at a temperature of t. s Most of the mixed water is recirculated back into the heat exchanger inside the dry cooling tower, where it is cooled by ambient temperatures of t. a The mass flow rate is W a The airflow cools the water, and the circulating water temperature drops to t. w Meanwhile, the airflow is heated to t0.

[0057] Define the cooling ratio m of the condenser as:

[0058]

[0059] In air-cooled systems, m is typically 30-50.

[0060] K is the heat transfer coefficient, and F is the calculated area. PL (KJ / (kg·K)) represents the specific heat capacity of cooling water at constant pressure; c pa (KJ / (kg·K)) represents the specific heat capacity of air at constant pressure.

[0061] The heat release Q1 from condensation in the air-cooled system is:

[0062] Q1 = D k r

[0063] Where r is the latent heat of vaporization of water.

[0064] The heat absorbed by the cooling water, Q2, is

[0065] Q2 = W L c PL (t s -t w )

[0066] The heat exchange capacity of the air-water heat exchanger is Q3.

[0067]

[0068] Where ε is the logarithmic mean temperature difference correction factor for the heat exchanger, and for cross-flow heat exchangers, ε is close to 1. Logarithmic mean temperature difference for:

[0069]

[0070] The heat absorbed by the air, Q4, is:

[0071] Q4 = W a c pa (t0-t a )

[0072] In a steady state, the energy balance must be maintained, thus:

[0073] Q1 = Q2 = Q3 = Q4

[0074] Solving the equations simultaneously, we get:

[0075]

[0076] Within the temperature range of 0-100℃, r is approximately equal to t. s Proportional, that is:

[0077] r = a - bt s

[0078] Substituting the above equation into the previous equation and rearranging, we get:

[0079]

[0080] Where W is about W a and W b The function β is defined as:

[0081] β = exp{NTU(W-1)}

[0082] Because a, b, c PL Both are constants, therefore the saturation temperature t of the condenser is... s It is t a The number of transfer units (NTU) is a function of m, W, and NTU, representing the ratio of the total thermal conductivity of the fluid to its heat capacity.

[0083] Within the temperature range of 0-100℃, the relationship between the saturation pressure of water vapor and the water temperature can be fitted by the following formula:

[0084]

[0085] When establishing a mechanistic model of a heat exchanger, assuming water or steam as the working fluid, all thermodynamically dense variables with respect to the x-axis and time t can be defined. Within this framework, the dynamic equilibrium equations for mass, momentum (ignoring the kinetic term), energy (ignoring the diffusion term), and the mass of each component can be formulated as follows:

[0086]

[0087]

[0088]

[0089]

[0090] Where A is the cross-sectional area of ​​the pipe, ρ is the fluid density, w is the mass flow rate, p is the fluid pressure, g is the acceleration due to gravity, z is the pipe height, and C is the fluid pressure. f ω is the Fanning coefficient of friction, h is the wetted perimeter, and φ is the specific enthalpy. e It is the heat flux entering the pipe through the side, ξ k N is the mass fraction of the k-th component. s This represents the amount of chemical substances in the fluid. The last equation must only consider multi-component, variable-component flow.

[0091] The dynamics of heat flow and temperature are described by the Fourier equations. By neglecting longitudinal heat conduction and concentrating the heat capacity in the middle of the tube thickness, a low-frequency approximation can be obtained, leading to the following one-dimensional model:

[0092] ρ m c m A m T m (x,t)=2πr i φ i (x,t)+2πr e φ e (x,t)

[0093] Where ρ m It is the density of the metal, c m It is the specific heat capacity of the metal, A m It is the cross-sectional area, T m It is the tube temperature in the middle of the thickness, r i and r e These are the inner and outer radii, φ, respectively. i and φ e This refers to the corresponding heat flow passing through the surface of the pipe; the latter two can be calculated as:

[0094]

[0095] Where λ is the thermal conductivity of the metal, T i and T e These are the temperatures of the inner and outer surfaces, respectively. Then, by sampling the above one-dimensional model at N equidistant nodes, it can be easily discretized.

[0096] When establishing the pump's mechanistic model, all pumps involved in the power system are variable speed pumps. The relationship between the pump's head H0 and flow rate G0 at its rated design speed can be expressed based on the pump's experimental performance curve, as follows:

[0097]

[0098] In the formula: K1, K2, and K3 are all coefficients, which can be obtained by fitting the pump's performance curve at the rated speed.

[0099] The relationship between the pump head H and flow rate G at any rotational speed n can be expressed by the pump similarity law as follows:

[0100] H = K1N 2 +K2NG+K3G 2

[0101]

[0102] In the formula: n is the pump speed, in r / min; N is the pump's nominal speed. At the rated speed n0, the pump efficiency η0 is:

[0103]

[0104] The relationship between pump efficiency n and flow rate G at any rotational speed n can be expressed by the pump similarity law as follows:

[0105]

[0106] The effective power of the pump and the power consumed by the steam turbine as external steam flows through the steam-driven pump are:

[0107]

[0108]

[0109]

[0110] In the formula: H is the pump head, in meters (m); w net The effective power of the pump is expressed in kilowatts (kW); W pump The turbine shaft power is the power of the pump when it reaches its effective power, in kilowatts (kW); ρ is the density of water, in kg / m³. 2 g is the acceleration due to gravity, with units of m / s². 3 p represents the boost pressure of the working fluid flowing through the pump, in MPa.

[0111] Pump outlet enthalpy:

[0112]

[0113] In the formula: h is the specific enthalpy of the working fluid flowing into the pump, in kJ / kg.

[0114] Based on the theorem of angular momentum and the principle of conservation of energy, a generalized Euler equation suitable for the start-up transient process of a centrifugal pump can be derived, as shown below.

[0115]

[0116] In the formula Ω J and Ω MThese are two different blade influence coefficients, the magnitude of which is related to impeller geometry parameters such as blade shape and blade thickness.

[0117] The mathematical model of the pump at different speeds is as follows:

[0118]

[0119] In the formula: H is the head at rotational speed n; Q is the flow rate at rotational speed n, in m³ / s. 3 / s;n d This is the pump's reference speed, also known as the design speed, and is expressed in r / min.

[0120] Then, step S2 is executed to configure input data for the input variables of the target component, determine the first measured value of the output variables of the target component, and input the input data into the mechanism model to obtain the first output value determined by the mechanism model. Taking a valve as an example, the input variable can be the pressure before the valve and the output variable can be the pressure after the valve. By giving the pressure before the valve as input data, the first measured value of the pressure after the valve can be obtained. Subsequently, the input data as the pressure before the valve is input into the mechanism model, and the first output value as the pressure after the valve can be calculated by the mechanism model.

[0121] Then, step S3 is executed to calculate the error between the first output value and the first measured value. This error is the error of the mechanism model.

[0122] The measurable parameters of the target component refer to the input parameters of the data-driven model. The second measured value of the measurable parameter can be obtained by measuring the target component in a real-world scenario using a measuring device. Subsequently, the second measured value of the measurable parameter of the target component and the error are input into the data-driven model, and the prediction error, i.e., the second output value, can be obtained through the data-driven model.

[0123] Finally, step S4 is executed. Using this second output value, error compensation can be performed on the first output value determined by the mechanism model to obtain the third output value. This third output value can then be used as the accurate output of the mechanism model, thereby correcting the output of the mechanism model and ensuring its accuracy.

[0124] The dynamic system hybrid modeling method provided in this embodiment of the invention first establishes a data-driven model and a mechanism model for the target component in the dynamic system. Then, it configures input data for the input variables of the target component, determines the first measured value of the output variables of the target component, and inputs the input data into the mechanism model to obtain the first output value determined by the mechanism model. Next, it calculates the error between the first output value and the first measured value, and inputs the second measured value of the measurable parameters of the target component and the error into the data-driven model to obtain the second output value of the data-driven model. Finally, based on the second output value, it performs error compensation on the first output value to obtain a third output value. This method compensates for the error in the output of the mechanism model through the data-driven model, reducing the dependence on the mechanism model. Even when the mechanism model is not clear, needs simplification, or is unknown, it can still obtain the accurate output of the mechanism model, thereby improving the overall accuracy and reliability of the modeling.

[0125] Based on the above embodiments, the dynamic system hybrid modeling method provided in this embodiment of the invention, wherein the mechanistic model and the data-driven model are connected in parallel and then in series with the mechanistic model; correspondingly, the step of performing error compensation on the first output value based on the second output value to obtain a third output value, and then including:

[0126] Based on the third output value, the mechanism model is optimized.

[0127] Specifically, such as Figure 4 As shown, the modeling structure of the target component can be achieved by connecting the mechanistic model and the data-driven model in parallel, followed by connecting them in series, thus realizing a hybrid parallel-serial modeling approach. Therefore, after error compensation for the first output value, the mechanistic model is optimized using the third output value. That is, the mechanistic parameters of the mechanistic model are adjusted using the third output value to obtain the optimized mechanistic model. Subsequently, only the optimized mechanistic model needs to be used; the unoptimized mechanistic model combined with error compensation is no longer required. This simplifies subsequent applications and improves the convenience of modeling applications.

[0128] Based on the above embodiments, the dynamic system hybrid modeling method provided in this embodiment of the invention involves connecting the data-driven model and the mechanism model in series, and then connecting them in parallel; correspondingly, the mechanism model is determined based on the following steps:

[0129] Based on the data-driven model, the mechanism model is established.

[0130] Specifically, such as Figure 5 As shown, the modeling structure of the target component can also be a series connection of the data-driven model and the mechanistic model, followed by a parallel connection of the data-driven model, to achieve a hybrid series-parallel modeling. Therefore, the mechanistic model can be established based on the data-driven model.

[0131] When establishing a mechanistic model, if the mechanism of the target component is unknown, the data-driven model can be directly used as the mechanistic model. If the mechanistic model of the target component is partially unknown, that is, there are mechanistic parameters with unknown values, a basic mechanistic model can be determined first. This basic mechanistic model is a mechanistic model that includes mechanistic parameters with unknown values, and the values ​​of the mechanistic parameters need to be determined with the help of a data-driven model.

[0132] The third measured value of the measurable parameter of the target component can be input into the data-driven model to obtain the fourth output value of the data-driven model. This third measured value can be the same as or different from the second measured value.

[0133] Subsequently, the values ​​of the mechanistic parameters of the mechanistic model can be determined using the third measurement value and the fourth output value. Then, the values ​​of the mechanistic parameters are used to update the mechanistic parameters with unknown values ​​in the basic mechanistic model, thus obtaining the mechanistic model.

[0134] Based on the above embodiments, the dynamic system hybrid modeling method provided in this embodiment of the invention obtains both the first and second measured values ​​based on one-dimensional mechanism simulation.

[0135] Specifically, in this embodiment of the invention, the first and second measured values ​​can be obtained either by measuring in a real-world scenario or by running a one-dimensional mechanism simulation. This reduces the measurement cost in a real-world scenario and allows the first and second measured values ​​to be obtained even in the absence of experimental data.

[0136] like Figure 6 As shown, based on the above embodiments, this embodiment of the invention provides a dynamic system hybrid modeling device, comprising:

[0137] Model building module 61 is used to build data-driven models and mechanism models of target components in a dynamic system;

[0138] The data preparation module 62 is used to configure input data for the input variables of the target component, determine the first measured value of the output variables of the target component, and input the input data into the mechanism model to obtain the first output value determined by the mechanism model.

[0139] Error calculation module 63 is used to calculate the error between the first output value and the first measured value, and input the second measured value of the measurable parameter of the target component and the error into the data-driven model to obtain the second output value of the data-driven model;

[0140] The error compensation module 64 is used to perform error compensation on the first output value based on the second output value to obtain a third output value.

[0141] Based on the above embodiments, the dynamic system hybrid modeling device provided in this embodiment of the invention includes a mechanism model and a data-driven model connected in parallel and then connected in series with the mechanism model; correspondingly, the dynamic system hybrid modeling device further includes an optimization module, used for:

[0142] Based on the third output value, the mechanism model is optimized.

[0143] Based on the above embodiments, the dynamic system hybrid modeling device provided in this embodiment of the invention, wherein the data-driven model is connected in series with the mechanism model and then connected in series with the mechanism model; correspondingly, the model building module is specifically used for:

[0144] Based on the data-driven model, the mechanism model is established.

[0145] Based on the above embodiments, the dynamic system hybrid modeling device provided in this embodiment of the invention further includes a model building module specifically used for:

[0146] The data-driven model is used as the mechanism model;

[0147] or,

[0148] A basic mechanism model is determined, and the third measured value of the measurable parameter of the target component is input into the data-driven model to obtain the fourth output value of the data-driven model;

[0149] Based on the third measurement value and the fourth output value, the values ​​of the mechanism parameters of the basic mechanism model are determined, and the mechanism model is constructed based on the values ​​of the mechanism parameters.

[0150] Based on the above embodiments, the dynamic system hybrid modeling device provided in this embodiment of the invention obtains the first and second measured values ​​based on one-dimensional mechanism simulation.

[0151] Based on the above embodiments, the dynamic system hybrid modeling device provided in this embodiment of the invention refers to a power system; the target component is one of a condenser, a heat exchanger, a pump, and a valve.

[0152] Based on the above embodiments, the dynamic system hybrid modeling device provided in the embodiments of the present invention has a data-driven model constructed based on a feedforward neural network or a long short-term memory neural network.

[0153] Specifically, the functions of each module in the dynamic system hybrid modeling device provided in this embodiment of the invention correspond one-to-one with the operation flow of each step in the above method-like embodiments, and the achieved effects are also the same. For details, please refer to the above embodiments, and this will not be repeated in this embodiment of the invention.

[0154] Figure 7 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 7 As shown, the electronic device may include a processor 710, a communications interface 720, a memory 730, and a communication bus 740, wherein the processor 710, the communications interface 720, and the memory 730 communicate with each other via the communication bus 740. The processor 710 can call logical instructions in the memory 730 to execute the dynamic system hybrid modeling method provided in the above embodiments.

[0155] Furthermore, the logical instructions in the aforementioned memory 730 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part 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 the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0156] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the dynamic system hybrid modeling method provided in the above embodiments.

[0157] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the dynamic system hybrid modeling method provided in the above embodiments.

[0158] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0159] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0160] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for hybrid modeling of dynamic systems, characterized in that, include: Establish data-driven and mechanistic models of target components in dynamic systems; Input data is configured for the input variables of the target component, a first measured value of the output variables of the target component is determined, and the input data is input into the mechanism model to obtain the first output value determined by the mechanism model; Calculate the error between the first output value and the first measured value, and input the second measured value of the measurable parameter of the target component and the error into the data-driven model to obtain the second output value of the data-driven model; Based on the second output value, error compensation is performed on the first output value to obtain the third output value; The mechanistic model and the data-driven model are connected in parallel and then connected in series with the mechanistic model; Accordingly, the step of performing error compensation on the first output value based on the second output value to obtain a third output value, and then includes: optimizing the mechanism model based on the third output value; Alternatively, the data-driven model and the mechanism model can be connected in series and then in parallel; correspondingly, the mechanism model is determined based on the following steps: establishing the mechanism model based on the data-driven model.

2. The dynamic system hybrid modeling method according to claim 1, characterized in that, The establishment of the mechanism model based on the data-driven model includes: The data-driven model is used as the mechanism model; or, A basic mechanism model is determined, and the third measured value of the measurable parameter of the target component is input into the data-driven model to obtain the fourth output value of the data-driven model; Based on the third measurement value and the fourth output value, the values ​​of the mechanism parameters of the basic mechanism model are determined, and the mechanism model is constructed based on the values ​​of the mechanism parameters.

3. The method for hybrid modeling of dynamic systems according to any one of claims 1-2, characterized in that, Both the first and second measured values ​​were obtained based on one-dimensional mechanism simulation.

4. The method for modeling hybrid dynamic systems according to any one of claims 1-2, characterized in that, The dynamic system is a power system; the target component is one of a condenser, a heat exchanger, a pump, and a valve.

5. The method for hybrid modeling of dynamic systems according to any one of claims 1-2, characterized in that, The data-driven model is constructed based on a feedforward neural network or a long short-term memory neural network.

6. A dynamic system hybrid modeling device, characterized in that, include: The model building module is used to build data-driven models and mechanism models of target components in dynamic systems; The data preparation module is used to configure input data for the input variables of the target component, determine the first measured value of the output variables of the target component, and input the input data into the mechanism model to obtain the first output value determined by the mechanism model. An error calculation module is used to calculate the error between the first output value and the first measured value, and input the second measured value of the measurable parameter of the target component and the error into the data-driven model to obtain the second output value of the data-driven model; An error compensation module is used to perform error compensation on the first output value based on the second output value to obtain a third output value; The mechanism model is connected in parallel with the data-driven model and then in series with the mechanism model; correspondingly, the dynamic system hybrid modeling device also includes an optimization module, used to: optimize the mechanism model based on the third output value; Alternatively, the data-driven model can be connected in series with the mechanistic model, and then connected in series with the mechanistic model; correspondingly, the model building module is specifically used to: build the mechanistic model based on the data-driven model.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the dynamic system hybrid modeling method as described in any one of claims 1-5.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the dynamic system hybrid modeling method as described in any one of claims 1-5.

Images