CN116070471A - A wind turbine simulation acceleration method and system based on reduced-order decomposition processing - Google Patents

Abstract

The invention discloses a wind driven generator simulation acceleration method and system based on reduced order decomposition processing, and belongs to the technical field of wind driven generator simulation data processing. The prior scheme does not disclose how to solve the problem of huge calculation consumption in the simulation process of the wind driven generator, and the wind driven generator simulation acceleration method based on the reduced order decomposition treatment carries out singular value decomposition on an initial data matrix, intercepts a plurality of singular values with larger energy occupation, and obtains a reduced order rank according to the quantity of the intercepted singular values; then according to the reduced rank, obtaining space-time characteristic quantity capable of decomposing the complex flow process into low rank; and then generating a reduced-order prediction matrix according to the space-time characteristic quantity, obtaining predicted flow field data of the time step to be calculated according to the reduced-order prediction matrix, realizing simulation acceleration of the wind driven generator, effectively reducing the calculation dimension and the calculation quantity, and effectively solving the problem of huge calculation consumption in the simulation process of the wind driven generator.

Description

A wind turbine simulation acceleration method and system based on reduced-order decomposition processing

Technical Field

The invention relates to a wind turbine simulation acceleration method and system based on reduced-order decomposition processing, belonging to the technical field of wind turbine simulation data processing.

Background Art

A Chinese patent (publication number CN114997078A) discloses a wind turbine flow field simulation test method and device, which includes: based on the three-dimensional flow field model of the wind turbine, performing transient fluid dynamics simulation on the wind turbine at the current moment, obtaining simulation data of the wind turbine and the three-dimensional flow field model at the current moment; based on the simulation data and the control strategy of the wind turbine, obtaining the simulation control parameters of the wind turbine at the next moment; based on the simulation control parameters, updating the reference coordinate system speed of the revolution area in the three-dimensional flow field model and the grid node speed of the rotation area in the three-dimensional flow field model. The wind turbine flow field simulation test method and device provided by the above scheme can perform non-steady-state computational fluid dynamics analysis on the wind rotor speed change and blade pitch change process when the wind turbine is in operation. The simulation takes into account the speed change and pitch change action of the wind turbine, and the transient flow field calculation results obtained are closer to the actual situation, and more accurate simulation test results can be obtained.

However, the above scheme does not disclose how to solve the problem of huge computational consumption in the simulation process of wind turbines. At present, the amount of data in the three-dimensional numerical simulation process is generally large, resulting in a long simulation calculation time. Especially for wind turbines, due to the limitations of the amount of grids and flow field data characteristics, the time consumed by their numerical simulation is often much longer than the physical time in reality, which will cause the existing wind turbine simulation technology to be unable to obtain the operating status of some key components of the wind turbine in a timely manner, and it will be impossible to know the operating status of key components in a timely manner, which will in turn affect the operation and maintenance of the wind turbine, and is not conducive to the promotion and utilization of wind turbine simulation technology.

Furthermore, due to the large amount of calculation and time consumption of wind turbine simulation technology, it will be difficult to directly apply wind turbine simulation technology to digital twins. Therefore, it is impossible to effectively build a benign interactive environment between the wind turbine model and the digital model, and it is impossible to form a technical system for digital regeneration, data synchronization, precise mapping and common progress of the wind turbine model and the digital model, and it is impossible to achieve full life cycle coverage of wind turbine simulation modeling.

Summary of the invention

In view of the defects of the prior art, the first purpose of the present invention is to provide a method for performing singular value decomposition on an initial data matrix to obtain several singular values, delete some singular values with a small energy proportion, and only intercept several singular values with a large energy proportion, and obtain a reduced order rank according to the number of intercepted singular values; then, according to the reduced order rank, the initial data matrix is subjected to dynamic modal reduced order decomposition to obtain time-space feature quantities that can decompose complex flow processes into low-rank; then, according to the time-space feature quantities, a reduced order prediction matrix is generated, and according to the reduced order prediction matrix, the predicted flow field data of the time step to be calculated is obtained, so as to realize the simulation acceleration of the wind turbine. The scheme is scientific and reasonable, and can effectively reduce the calculation dimension, reduce the amount of calculation, save calculation time and server calculation load. A wind turbine simulation acceleration method based on reduced order decomposition processing.

The second purpose of the present invention is to provide a method for accelerating the simulation of wind turbines by performing a reduced-order decomposition process on initial flow field data to obtain low-rank spatiotemporal feature quantities, and predicting the flow field data of future time steps based on the spatiotemporal feature quantities, thereby effectively reducing the amount of data processing and the simulation calculation time, thereby effectively solving the problem of huge computing consumption in the simulation process of wind turbines; thereby facilitating users to obtain the operating status of some key components of wind turbines in a timely manner, and promptly knowing the operating status of key components, which can effectively improve the operation and maintenance efficiency of wind turbines and is conducive to the promotion and utilization of wind turbine simulation technology.

The third object of the present invention is to provide a wind turbine simulation acceleration method which realizes simulation acceleration of wind turbines by constructing a reduced-order calculation model and a data prediction reduced-order model, which can effectively reduce the calculation dimension, reduce the calculation amount, and save calculation time.

The fourth objective of the present invention is to provide a wind turbine simulation acceleration method and system that has a small amount of data processing, short time consumption, can effectively accelerate calculations, and has high accuracy, so that numerical simulation technology can be more suitable for digital twin models.

To achieve one of the above purposes, the first technical solution of the present invention is:

A wind turbine simulation acceleration method based on reduced-order decomposition processing comprises the following steps:

The first step is to perform full-order fluid dynamics simulation calculations on the wind turbine to obtain the full-order initial flow field data at a certain time step;

The second step is to process the initial flow field data and construct the initial data matrix;

The third step is to perform singular value decomposition on the initial data matrix to obtain several singular values, and determine the reduced rank according to the energy proportion of the singular values;

The fourth step is to perform dynamic modal reduction decomposition on the initial data matrix according to the reduced rank to obtain the spatiotemporal characteristic quantities used to characterize the flow process of the flow field;

The fifth step is to generate a reduced-order prediction matrix according to the spatiotemporal feature quantity, and obtain the predicted flow field data of the time step to be calculated according to the reduced-order prediction matrix;

The sixth step is to simulate and calculate the wind turbine based on the predicted flow field data to achieve simulation acceleration of the wind turbine.

After continuous exploration and experimentation, the present invention performs singular value decomposition on the initial data matrix to obtain several singular values. According to the principle that the larger the singular value, the greater the contribution to the entire matrix, some singular values with a small energy share are deleted, and only some singular values with a large energy share are intercepted, thereby reducing the impact on the accuracy of simulation calculation while reducing the amount of data processing. And according to the number of intercepted singular values, the reduced order rank is obtained; then according to the reduced order rank, the initial data matrix is subjected to dynamic modal reduced order decomposition to obtain a time-space feature quantity that can decompose a complex flow process into a low-rank; then according to the time-space feature quantity, a reduced order prediction matrix is generated, and finally according to the reduced order prediction matrix, the predicted flow field data of the time step to be calculated is obtained, so as to realize the simulation acceleration of the wind turbine. The scheme of the present invention is scientific and reasonable, and can effectively reduce the calculation dimension, reduce the amount of calculation, save calculation time and server calculation load, and thus can effectively solve the problem of huge calculation consumption in the simulation process of the wind turbine.

Furthermore, the present invention obtains low-rank spatiotemporal characteristics by performing a reduced-order decomposition process on the initial flow field data, and predicts the flow field data of future time steps based on the spatiotemporal characteristics, thereby effectively reducing the amount of data processing and the simulation calculation time, thereby effectively solving the problem of huge computing consumption in the simulation process of wind turbines; and thus facilitating users to obtain the operating status of some key components of wind turbines in a timely manner, and timely knowing the operating status of key components, which can effectively improve the operation and maintenance efficiency of wind turbines and is conducive to the promotion and utilization of wind turbine simulation technology.

Furthermore, compared with the full-order fluid mechanics simulation calculations in the prior art, the method for predicting future flow field data of the present invention takes less time, can effectively accelerate the calculation, and has a very high accuracy, making the numerical simulation technology more suitable for the digital twin model. Therefore, the simulation acceleration method of the present invention can effectively construct a benign interactive environment between the wind turbine model and the digital model, forming a technical system for digital regeneration, data synchronization, precise mapping and common progress of the wind turbine model and the digital model, thereby achieving full life cycle coverage of wind turbine simulation modeling. The solution is scientific, reasonable and feasible.

As the preferred technical measures:

In the first step, the method for performing full-order fluid dynamics simulation calculation on the wind turbine is as follows:

Step 11, constructing a geometric model of the wind turbine;

Step 12, based on the geometric model, setting meteorological conditions and working conditions of the wind turbine generator to obtain a calculation grid;

Step 13, based on computational fluid dynamics, perform full-order flow field calculation on the computational grid to obtain initial flow field data including multiple time steps.

As the preferred technical measures:

Meteorological conditions include wind speed and direction;

The working condition is the wind turbine speed;

The initial flow field data include physical quantities on each computational grid;

The physical quantity at least includes speed and/or pressure and/or temperature.

As the preferred technical measures:

In the second step, the method of constructing the initial data matrix is as follows:

Step 21, obtaining the physical quantity on each calculation grid;

Step 22, storing the retrieved physical quantity into a digital matrix as required to form an initial data matrix;

The first column of the initial data matrix is the physical quantity of the first time step, and the last column is the physical quantity of the N-1th time step;

The i -th row of the initial data matrix represents the process information of the physical quantity on the i-th computational grid changing with time, and its j -th column represents the physical quantity at the j-th time step.

As the preferred technical measures:

In the third step, the method of performing singular value decomposition on the initial data matrix is as follows:

Step 31, performing a singular value decomposition operation on the initial data matrix to obtain a singular value matrix;

Step 32, setting the singular values arranged from large to small on the diagonal of the singular value matrix;

Step 33, obtaining the energy contribution of each singular value to the initial data matrix;

Step 32, accumulating energy contributions one by one from large to small to obtain the energy contribution sum. When the energy contribution sum is greater than the energy threshold, the number of energy contributions participating in the accumulation is counted, and the number of energy contributions is used as the reduced rank.

As the preferred technical measures:

In the fourth step, the method of performing dynamic mode reduction decomposition on the initial data matrix is as follows:

Step 41, constructing an initial prediction matrix with a time step difference according to the initial data matrix;

Step 42, performing truncated singular value decomposition on the initial flow field data according to the reduced order to obtain a decomposition result;

Step 43, constructing a decomposition similarity matrix between the initial flow field data and the initial prediction matrix according to the decomposition result and the initial prediction matrix;

Step 44, performing eigenvalue decomposition on the decomposed similarity matrix to obtain eigenvalues and eigenvectors;

Step 45, constructing a spatiotemporal characteristic quantity for characterizing the flow process of the wind turbine generator according to the decomposition result, the initial prediction matrix, the eigenvalues and the eigenvectors.

As the preferred technical measures:

The method to construct the initial prediction matrix is as follows:

The first column of the initial prediction matrix is the flow field data of the second time step, and the last column is the flow field data of the last time step;

The i- th row of the initial prediction matrix represents the process information of the physical quantity on the i-th computational grid changing with time, and its j -th column represents the physical quantity at the j+1-th time step.

As the preferred technical measures:

In the fifth step, the method for obtaining the predicted flow field data of the time step to be calculated is as follows:

Step 51, constructing a reduced-order prediction matrix according to the spatiotemporal feature quantity, the initial flow field data and the initial prediction matrix, wherein the reduced-order prediction matrix includes the flow field data of the time step to be calculated;

Step 52, calculating the reduced-order prediction matrix to obtain the predicted flow field data of the time step to be calculated;

Step 53, adding the predicted flow field data of the time step to be calculated to the initial prediction matrix and the initial flow field data, and constructing a new initial prediction matrix and a new initial data matrix;

Step 54, based on the new initial prediction matrix and the new initial data matrix, obtain the predicted flow field data that is one time step different from the new initial prediction matrix, repeat the cycle, complete the calculation of the flow field data for several time steps to be calculated, and realize the future prediction of the flow field.

To achieve one of the above purposes, the second technical solution of the present invention is:

A wind turbine simulation acceleration method based on reduced-order decomposition processing includes the following contents:

Using the pre-built fluid mechanics simulation model, a full-order fluid mechanics simulation calculation is performed on the wind turbine to obtain the full-order initial flow field data at a certain time step;

The initial flow field data is processed through a pre-built reduced-order calculation model to construct an initial data matrix; singular value decomposition is performed on the initial data matrix to obtain a number of singular values, and the reduced-order rank is determined based on the energy proportion of the singular values;

Using the pre-built data prediction reduction model, and according to the reduction rank, the initial data matrix is subjected to dynamic modal reduction decomposition to obtain the spatiotemporal characteristic quantities used to characterize the flow process of the flow field; then, based on the spatiotemporal characteristic quantities, a reduction prediction matrix is generated, and based on the reduction prediction matrix, the predicted flow field data of the time step to be calculated is obtained;

Through the reduced-order calculation model and the data prediction reduced-order model, the predicted flow field data is obtained, the wind turbine is simulated and calculated, and the simulation acceleration of the wind turbine is achieved.

After continuous exploration and experimentation, the present invention constructs a reduced-order calculation model, performs singular value decomposition on the initial data matrix, obtains several singular values, and deletes some singular values with a small energy proportion according to the principle that the larger the singular value, the greater the contribution to the entire matrix, and only intercepts several singular values with a large energy proportion, thereby reducing the data processing amount and reducing the impact on the accuracy of the simulation calculation. And according to the number of intercepted singular values, the reduced-order rank is obtained; then the data prediction reduced-order model is used, and according to the reduced-order rank, the initial data matrix is subjected to dynamic modal reduced-order decomposition to obtain a time-space feature quantity that can decompose a complex flow process into a low-rank; then according to the time-space feature quantity, a reduced-order prediction matrix is generated, and according to the reduced-order prediction matrix, the predicted flow field data of the time step to be calculated is obtained, so as to realize the simulation acceleration of the wind turbine. The scheme of the present invention is scientific and reasonable, and can effectively reduce the calculation dimension, reduce the calculation amount, save calculation time and server calculation load.

Furthermore, the present invention obtains low-rank spatiotemporal characteristics by performing a reduced-order decomposition process on the initial flow field data, and predicts the flow field data of future time steps based on the spatiotemporal characteristics, thereby effectively reducing the amount of data processing and the simulation calculation time, thereby effectively solving the problem of huge computing consumption in the simulation process of wind turbines; and thus facilitating users to obtain the operating status of some key components of wind turbines in a timely manner, and timely knowing the operating status of key components, which can effectively improve the operation and maintenance efficiency of wind turbines and is conducive to the promotion and utilization of wind turbine simulation technology.

Furthermore, compared with the full-order fluid mechanics simulation calculations in the prior art, the method of predicting future flow field data in the present invention takes less time, can effectively accelerate the calculation, and has a high degree of accuracy, making numerical simulation technology more suitable for digital twin models.

To achieve one of the above purposes, the third technical solution of the present invention is:

A wind turbine simulation acceleration system based on reduced-order decomposition processing, comprising:

one or more processors;

A storage device for storing one or more programs;

When the one or more programs are executed by the one or more processors, the one or more processors implement the above-mentioned wind turbine simulation acceleration method based on reduced-order decomposition processing.

Compared with the prior art, the present invention has the following beneficial effects:

After continuous exploration and experimentation, the present invention performs singular value decomposition on the initial data matrix to obtain several singular values. According to the principle that the larger the singular value, the greater the contribution to the entire matrix, some singular values with a small energy share are deleted, and only some singular values with a large energy share are intercepted, thereby reducing the impact on the accuracy of simulation calculation while reducing the amount of data processing. And according to the number of intercepted singular values, the reduced order rank is obtained; then according to the reduced order rank, the initial data matrix is subjected to dynamic modal reduced order decomposition to obtain a time-space feature quantity that can decompose a complex flow process into a low-rank; then according to the time-space feature quantity, a reduced order prediction matrix is generated, and according to the reduced order prediction matrix, the predicted flow field data of the time step to be calculated is obtained, so as to realize the simulation acceleration of the wind turbine. The scheme is scientific and reasonable, and can effectively reduce the calculation dimension, reduce the amount of calculation, save calculation time and server calculation load.

Furthermore, the present invention obtains low-rank spatiotemporal characteristics by performing a reduced-order decomposition process on the initial flow field data, and predicts the flow field data of future time steps based on the spatiotemporal characteristics, thereby effectively reducing the amount of data processing and the simulation calculation time, thereby effectively solving the problem of huge computing consumption in the simulation process of wind turbines; and thus facilitating users to obtain the operating status of some key components of wind turbines in a timely manner, and timely knowing the operating status of key components, which can effectively improve the operation and maintenance efficiency of wind turbines and is conducive to the promotion and utilization of wind turbine simulation technology.

Furthermore, compared with the full-order fluid mechanics simulation calculations in the prior art, the method of predicting future flow field data in the present invention takes less time, can effectively accelerate the calculation, and has a high degree of accuracy, making numerical simulation technology more suitable for digital twin models.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic flow chart of a wind turbine simulation acceleration method according to the present invention;

FIG2 is another schematic flow chart of the wind turbine generator simulation acceleration method of the present invention;

FIG3 is a schematic diagram of a corresponding relationship between singular values and energy contributions of the present invention;

FIG4 is a schematic diagram showing a changing relationship between the singular value order and the energy contribution sum of the present invention;

FIG5 is a diagram showing the variation of the maximum error and the average error with the prediction time step according to the present invention.

DETAILED DESCRIPTION

In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

On the contrary, the present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention as defined by the claims. Further, in order to make the public have a better understanding of the present invention, some specific details are described in detail in the detailed description of the present invention below. Those skilled in the art can fully understand the present invention without the description of these details.

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

As shown in FIG1 , a first specific embodiment of the wind turbine simulation acceleration method based on reduced-order decomposition processing of the present invention is as follows:

A wind turbine simulation acceleration method based on reduced-order decomposition processing comprises the following steps:

The first step is to perform full-order fluid dynamics simulation calculations on the wind turbine to obtain the full-order initial flow field data at a certain time step;

The second step is to process the initial flow field data and construct the initial data matrix;

The third step is to perform singular value decomposition on the initial data matrix to obtain several singular values, and determine the reduced rank according to the energy proportion of the singular values;

The fourth step is to perform dynamic modal reduction decomposition on the initial data matrix according to the reduced rank to obtain the spatiotemporal characteristic quantities used to characterize the flow process of the flow field;

The fifth step is to generate a reduced-order prediction matrix according to the spatiotemporal feature quantity, and obtain the predicted flow field data of the time step to be calculated according to the reduced-order prediction matrix;

The sixth step is to simulate and calculate the wind turbine based on the predicted flow field data to achieve simulation acceleration of the wind turbine.

A second specific embodiment of the wind turbine simulation acceleration method based on reduced-order decomposition processing of the present invention:

A wind turbine simulation acceleration method based on reduced-order decomposition processing comprises the following steps:

The first step is to perform full-order fluid dynamics simulation calculations on the wind turbine to obtain the full-order initial flow field data at a certain time step;

The second step is to process the initial flow field data and construct an initial data matrix and an initial prediction matrix with a difference of one or more time steps;

The third step is to perform singular value decomposition on the initial data matrix to obtain several singular values, and determine the reduced rank according to the energy proportion of the singular values;

The fourth step is to perform dynamic mode decomposition on the initial data matrix and the initial prediction matrix according to the reduced rank to determine the approximate matrix between the two;

The fifth step is to perform eigendecomposition on the approximate matrix to obtain the spatiotemporal characteristic quantities used to characterize the flow process of the wind turbine flow field;

The sixth step is to generate a reduced-order prediction matrix according to the spatiotemporal feature quantities, and obtain the predicted flow field data for the next time step or several time steps according to the reduced-order prediction matrix;

The seventh step is to simulate and calculate the wind turbine according to the predicted flow field data to achieve simulation acceleration of the wind turbine.

A third specific embodiment of the wind turbine simulation acceleration method based on reduced-order decomposition processing of the present invention:

A wind turbine simulation acceleration method based on reduced-order decomposition processing comprises the following steps:

The first step is to use the pre-built fluid mechanics simulation model to perform full-order fluid mechanics simulation calculations on the wind turbine to obtain full-order flow field data at a certain time step;

The second step is to process the flow field data through the pre-built reduced-order calculation model to obtain the initial data matrix, and then perform singular value decomposition on the initial data matrix to obtain several singular values, and determine the reduced-order rank of the initial data matrix based on the energy proportion of the singular values;

The third step is to process the flow field data according to the pre-built data prediction reduction model to obtain the data prediction matrix, and then truncate and decompose the data prediction matrix according to the reduced order rank to obtain the spatiotemporal characteristic quantities used to characterize the flow process of the wind turbine; reconstruct the data prediction matrix according to the spatiotemporal characteristic quantities to generate a reduced order prediction matrix, which includes the flow field data of the time step to be calculated; and obtain the predicted flow field data of the next time step or several time steps according to the reduced order prediction matrix;

The fourth step is to simulate and accelerate the wind turbine based on the predicted flow field data.

A fourth specific embodiment of the wind turbine simulation acceleration method based on reduced-order decomposition processing of the present invention:

A wind turbine simulation acceleration method based on reduced-order decomposition processing includes a computational fluid dynamics numerical solution model, a singular value decomposition model and a data dynamic mode decomposition model.

The computational fluid dynamics numerical solution model is used to calculate the flow field of the wind turbine in the first N time steps. It replaces the integral or partial differential equation in the control equation with a discrete algebraic form, and then solves the specific numerical solution for data comparison and explanation of experimental phenomena. The physical basis behind it and the control equations based on different physical laws are listed below. The control equations derived from three physical laws are as follows:

Conservation of matter:

Conservation of Momentum:

Conservation of Energy:

为密度，

是时间，

为速度，p为压强，

为剪切力，

为温度，

为热导率，

为等压比热容，

为加热/冷却源项。In the formula,

is the density,

It's time.

is the velocity, p is the pressure,

is the shear force,

is the temperature,

is the thermal conductivity,

is the isobaric specific heat capacity,

is the heating/cooling source term.

The singular value decomposition model is used to solve the flow modes of the flow field and the energy contained in each mode. The interception rank r of the entire flow field can be characterized according to the energy interception of the mode.

，可以将它分解为一个正定矩阵

，一个对角矩阵

，以及另一个正定矩阵的转置

的乘积。For the first N steps of raw data matrix obtained from the computational fluid dynamics simulation

, which can be decomposed into a positive definite matrix

, a diagonal matrix

, and the transpose of another positive definite matrix

The product of .

的表达式如下所示：The original data matrix of the first N steps

The expression is as follows:

。

.

The expression for the product is as follows:

所代表的线性变换可以由更简单的旋转，拉伸变换进行合成。Product Expression Represents Matrix

The represented linear transformations can be synthesized from simpler rotations and stretches.

为奇异值矩阵，对角线上分布着奇异值，由大到小排列。奇异值越大，对整个矩阵的贡献就越大，根据这个原理，可以删除一些较小的奇异值，截取前r项奇异值，仅保留较大的奇异值，通过前r项奇异值计算近似矩阵

，就可以近似得到原始数据矩阵，原始数据矩阵的表达式如下所示：Diagonal matrix in singular value decomposition model

It is a singular value matrix, with singular values distributed on the diagonal, arranged from large to small. The larger the singular value, the greater its contribution to the entire matrix. According to this principle, some smaller singular values can be deleted, the first r singular values can be intercepted, and only the larger singular values can be retained. The approximate matrix can be calculated through the first r singular values.

, we can approximate the original data matrix, and the expression of the original data matrix is as follows:

，

，

，分别为

、

、

的截断矩阵。Among them, the matrix

,

,

, respectively

,

,

The truncated matrix of .

The data dynamic mode decomposition model is a data-driven method that can be used to analyze the dynamic process of fluids (such as water flow) and can decompose complex flow processes into low-rank spatiotemporal features. Dynamic mode decomposition has many advantages in describing some dynamic processes, including not relying on any given dynamic system expression, and being able to make short-term state predictions. The model itself has predictive capabilities.

，x为一维数据上每个点的数值，下标表示空间采样点，数量为M个；t为整个数据随时间的变化，下标表示时间采样点，数量为N个；数据矩阵中的物理量（速度、温度、浓度等）为u，例如

就表示第N个时间步时，第M个网格点上的物理量，数据矩阵

的表达式如下所示：The basic idea of dynamic mode decomposition is linear transformation. For a data matrix

, x is the value of each point on the one-dimensional data, the subscript represents the spatial sampling point, the number is M; t is the change of the entire data over time, the subscript represents the time sampling point, the number is N; the physical quantity (speed, temperature, concentration, etc.) in the data matrix is u , for example

It means the physical quantity at the Mth grid point at the Nth time step, the data matrix

The expression is as follows:

，其表达式如下所示：At the original time t = 1, the data can be recorded in the form of a column vector as

, whose expression is as follows:

，使得：In the data dynamic mode decomposition model, if the system is a linear system, a matrix can be found

, so that:

，以及系统的变换矩阵

，就可以知道系统之后任意时刻t状态

。对于非线性系统，也可以找到近似的矩阵

，以进行数据降维处理，但会产生一定的误差

，其表达式如下：So as long as you know the original state

, and the transformation matrix of the system

, we can know the state of the system at any time t

For nonlinear systems, we can also find an approximate matrix

, to reduce the dimension of data, but it will produce certain errors

, which is expressed as follows:

Construct a data prediction matrix based on the dynamic mode decomposition of the first N steps read.

为第1步至第N-1步组成的矩阵，其表达式如下：matrix

is the matrix composed of steps 1 to N-1, and its expression is as follows:

为第2步至第N步组成的矩阵，其表达式如下：matrix

is the matrix composed of steps 2 to N, and its expression is as follows:

。but

.

为库普曼矩阵，可以使用低秩结构进行逼近。in,

is a Koopman matrix and can be approximated using a low-rank structure.

的方法如下：In the data dynamic mode decomposition model, solving the Koopman matrix

The method is as follows:

的截断奇异值分解，其表达式如下：According to the cutoff rank r of the singular value decomposition, calculate

The truncated singular value decomposition of is expressed as follows:

Use the following matrix:

进行近似求解，其中矩阵

，

，

，分别为

的

、

、

截断矩阵，矩阵

。Koopman Matrix

Perform an approximate solution, where the matrix

,

,

, respectively

of

,

,

Truncated matrix, matrix

.

进行特征值分解，其计算公式如下：Approximate Koopman Matrix

Perform eigenvalue decomposition, and the calculation formula is as follows:

为对角线元素为特征值的对角矩阵，矩阵

由特征向量构成，可以用特征值和特征向量来分析复杂流动过程的时空特征。在动态模态分解中，构建分解矩阵，其表达式如下所示：The spatiotemporal feature analysis is performed with the help of dynamic mode decomposition, where

is a diagonal matrix whose diagonal elements are eigenvalues, and the matrix

It is composed of eigenvectors, and eigenvalues and eigenvectors can be used to analyze the spatiotemporal characteristics of complex flow processes. In dynamic mode decomposition, a decomposition matrix is constructed, and its expression is as follows:

；

;

进行计算，其公式如下所示：According to the decomposition matrix, the short-term prediction matrix including the next time step is

The calculation formula is as follows:

为矩阵

的伪逆矩阵，伪逆矩阵是逆矩阵的广义形式，由于奇异矩阵或非方阵的矩阵，例如

，不存在逆矩阵，因此通过一般化的逆矩阵，即伪逆矩阵来解决这个问题。in,

For the matrix

The pseudo-inverse matrix of the inverse matrix is a generalized form of the inverse matrix. Due to the singular matrix or non-square matrix, for example

, there is no inverse matrix, so this problem is solved by a generalized inverse matrix, namely the pseudo-inverse matrix.

，

的伪逆矩阵

必然存在，且

满足以下条件：For any matrix

,

The pseudo-inverse matrix

must exist, and

The following conditions are met:

是一个效果等同单位矩阵I，但又不是单位矩阵I的矩阵。So the matrix

is a matrix that is equivalent to the identity matrix I , but is not the identity matrix I.

和

，形成新的预测矩阵

和

，

。After predicting the N+1th step, add the predicted physical field data to the prediction matrix

and

, forming a new prediction matrix

and

,

.

为第1步至第N步组成的矩阵，其表达式如下所示：New prediction matrix

is the matrix composed of step 1 to step N, and its expression is as follows:

为第2步至第N+1步组成的矩阵，其表达式如下所示：New prediction matrix

is the matrix composed of steps 2 to N+1, and its expression is as follows:

和

继续预测第N+2步的流场数据，如此迭代循环，实现对未来流场的预测。Using the same method, according to

and

Continue to predict the flow field data of step N+2, and repeat this iterative cycle to achieve the prediction of future flow fields.

Compared with the original computational fluid dynamics simulation calculation, the method of future prediction through reduced-order model in the present invention takes less time, can accelerate the calculation extremely quickly, has the advantages of high efficiency, high accuracy, small error, etc., and can solve the problems of high consumption and long calculation time of simulation calculation.

As shown in FIG. 2 , the fifth embodiment of the wind turbine simulation acceleration method based on reduced-order decomposition processing of the present invention is as follows:

A wind turbine simulation acceleration method based on reduced-order decomposition processing comprises the following steps:

Step 1: Perform computational fluid dynamics simulation calculations to obtain flow field data for N time steps;

；Step 2: Construct the original data matrix through N-step flow field data

;

进行奇异值分解，根据每个奇异值所对应的能量，选取截断秩r，使得降阶之后的模型能够涵盖流场中绝大多数信息；Step 3: The original data matrix

Perform singular value decomposition and select the cutoff rank r according to the energy corresponding to each singular value, so that the reduced-order model can cover most of the information in the flow field;

、

，预测数据的矩阵的行表示网格量，每一列为一个时间步的数据，其中

包括前N-1个时间步的流场数据，

包括第2个至第N个时间步的流场数据：Step 4: Construct a prediction data matrix using N-step flow field data

,

, the rows of the matrix of predicted data represent the grid quantity, and each column represents the data of one time step, where

Including the flow field data of the first N-1 time steps,

Includes flow field data from the 2nd to the Nth time step:

、

进行动态模态分解处理，得到库普曼矩阵

，并近似求解库普曼矩阵

，库普曼矩阵

的计算公式如下所示：Step 5: Based on the cutoff rank r calculated in step 3 and the predicted data matrix in step 4

,

Perform dynamic mode decomposition to obtain the Koopman matrix

, and approximately solve the Koopman matrix

, Koopman matrix

The calculation formula is as follows:

，

,

、

、

为矩阵

的截断奇异值分解结果。The matrix

,

,

For the matrix

The truncated singular value decomposition result of .

进行特征值分解：

，使用矩阵

的特征值和特征向量来分析复杂流动过程的时空特征，其表达式如下所示：Then the Koopman matrix

Perform eigenvalue decomposition:

, using the matrix

The eigenvalues and eigenvectors are used to analyze the spatiotemporal characteristics of complex flow processes, and the expressions are as follows:

可以计算为：

，矩阵

的最后一列为未计算的N+1时间步的流场数据；Short-term forecast matrix

It can be calculated as:

,matrix

The last column is the flow field data of the uncalculated N+1 time step;

和

，构建新的数据预测矩阵

和原始数据矩阵

，同时将矩阵

拓展至第N步流场，形成新的矩阵

，其表达式如下所示：Step 6: Add the predicted N+1th time step data to the matrix

and

, construct a new data prediction matrix

and the original data matrix

, and the matrix

Expand to the Nth step flow field to form a new matrix

, whose expression is as follows:

Repeat steps 3 to 6 to achieve short-term prediction of future flow fields.

An acceleration method for wind turbine simulation based on reduced-order decomposition processing provided by an embodiment of the present invention can perform short-term prediction of flow field based on existing computational fluid dynamics calculation data, thereby saving calculation time.

A specific embodiment of applying the present invention to accelerate the temperature field calculation of a wind turbine nacelle:

In wind power operation and maintenance, due to long-term sunshine and poor ventilation in hot weather, heat dissipation is difficult, and it is very easy to generate local high temperatures in the cabin, causing damage to some electrical components and even causing serious accidents such as fires. Therefore, in the digital twin, the development of the temperature field in the cabin in the future period of time can be predicted in time according to meteorological conditions. The operation and maintenance strategy of the wind turbine can be adjusted in time, and relevant adjustments can be made to avoid damage to electronic components or serious accidents. At the same time, in order to reduce the amount of calculation and simulation time, the temperature field calculation of the wind turbine cabin needs to be accelerated.

The specific implementation process of applying the present invention to accelerate the temperature field calculation of the wind turbine nacelle is as follows:

Step 1: Process the nacelle of a certain type of wind turbine to obtain a simplified geometric model. Since the influence of some details in the nacelle on the flow field can be ignored, the internal geometry is simplified to the main heat-generating components such as the generator, electrical cabinet, gearbox and housing in the computational fluid dynamics simulation calculation;

According to the simplified geometric model, the computational fluid dynamics grid model of a certain type of wind turbine nacelle is obtained, with a grid volume of 471184 and a structured grid type.

Step 2: Calculate the temperature rise of a certain type of wind turbine cabin under meteorological conditions of 40°C temperature and 3m/s wind speed. A total of 200 seconds of transient flow field are calculated with a time step of 1 second. The temperature field of the first 100 seconds is selected as training data, and the temperature field of the last 100 seconds is selected as verification data.

，共计471184行，100列数据，

。Constructing the original data matrix

, a total of 471184 rows and 100 columns of data,

.

进行奇异值分解，分别计算每阶奇异值所对应的能量贡献，可参见图3；计算前N阶奇异值对应的能量贡献和，并截取具有99%以上能量贡献累积的阶数，可参见图4。本实施例中，前11阶奇异值贡献了99.79%的能量，因此取截断秩

。Step 3: Original data matrix

Perform singular value decomposition, calculate the energy contribution corresponding to each order singular value, see Figure 3; calculate the sum of the energy contributions corresponding to the first N order singular values, and cut off the order with more than 99% energy contribution accumulation, see Figure 4. In this embodiment, the first 11 order singular values contribute 99.79% of the energy, so the cutoff order is taken

.

及

，

，其表达式如下所示：Step 4: Construct data prediction matrix

and

,

, whose expression is as follows:

的截断奇异值分解结果

、

、

：

；求解库普曼矩阵

，

，对矩阵

进行特征值分解：

。Step 5: Predict the temperature field in the future time step through the data dynamic mode decomposition model. First, solve the matrix

The truncated singular value decomposition result of

,

,

:

; Solve for the Koopman matrix

,

, for the matrix

Perform eigenvalue decomposition:

.

的特征值和特征向量计算最后一列为第101个时间步温度数据的短期预测矩阵

，其表达式如下所示：Through the matrix

The last column of the eigenvalue and eigenvector calculation is the short-term prediction matrix of the temperature data at the 101st time step.

, whose expression is as follows:

最后一列加入至矩阵

、

、

，构建新的矩阵

'、

和

，

，

，其表达式如下所示：Step 6: The matrix

The last column is added to the matrix

,

,

, construct a new matrix

'、

and

,

,

, whose expression is as follows:

'、

和

重复步骤3~5，预测第102步的温度数据，重复迭代此流程，实现对未来流场数据的预测，实现计算加速。Using the Matrix

'、

and

Repeat steps 3 to 5 to predict the temperature data of step 102. Repeat this process to predict future flow field data and achieve computational acceleration.

After experiments, a workstation with an 8-core i7-9700 processor and 8G memory was used to perform computational fluid dynamics simulation on the above scenario, and each step of calculation took 5 minutes. However, the wind turbine simulation acceleration method based on reduced-order decomposition processing of the present invention can accelerate the calculation time of each time step to about 5 seconds.

As shown in FIG5 , although the maximum error [K] and the average error [K] of this embodiment may vary with the prediction time step, the error of this embodiment can be controlled within 5%.

Therefore, the present invention is applied to accelerate the temperature field calculation of the wind turbine nacelle, which can accelerate the original computational fluid dynamics calculation by 60 times and still have good accuracy.

An embodiment of a device applying the method of the present invention:

A computer device comprising:

one or more processors;

A storage device for storing one or more programs;

When the one or more programs are executed by the one or more processors, the one or more processors implement the above-mentioned wind turbine simulation acceleration method based on reduced-order decomposition processing.

A computer medium embodiment of the method of the present invention is applied:

A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above-mentioned wind turbine simulation acceleration method based on reduced-order decomposition processing.

Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, and computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

The present application is described with reference to the flowcharts and/or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and/or box in the flowchart and/or block diagram and the combination of the processes and/or boxes in the flowchart and/or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate a device for implementing the functions specified in one process or multiple processes in the flowchart and/or one box or multiple boxes in the block diagram.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device that implements the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.

These computer program instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims (10)

1. A wind driven generator simulation acceleration method based on reduced order decomposition processing is characterized in that,

the method comprises the following steps:

firstly, performing full-order hydrodynamic simulation calculation on a wind driven generator to obtain full-order initial flow field data of a certain time step;

secondly, processing the initial flow field data to construct an initial data matrix;

thirdly, performing singular value decomposition on the initial data matrix to obtain a plurality of singular values, and determining a reduced rank according to the energy duty ratio of the singular values;

fourthly, performing dynamic modal reduced order decomposition on the initial data matrix according to the reduced order rank to obtain space-time characteristic quantity used for representing the flow process of the flow field;

fifthly, generating a reduced-order prediction matrix according to the space-time characteristic quantity, and obtaining predicted flow field data of the time step to be calculated according to the reduced-order prediction matrix;

and sixthly, carrying out simulation calculation on the wind driven generator according to the predicted flow field data, and realizing simulation acceleration of the wind driven generator.

2. The wind driven generator simulation acceleration method based on reduced order decomposition processing according to claim 1, wherein,

in the first step, the method for performing full-order hydrodynamic simulation calculation on the wind driven generator comprises the following steps:

step 11, constructing a geometric model of the wind driven generator;

step 12, setting meteorological conditions and working conditions of a wind driven generator on the basis of a geometric model to obtain a calculation grid;

and 13, performing full-order flow field calculation on the calculation grid according to computational fluid dynamics to obtain initial flow field data comprising a plurality of time steps.

3. A wind driven generator simulation acceleration method based on reduced order decomposition processing as set forth in claim 2, wherein,

meteorological conditions include wind speed and wind direction;

the working condition is the rotating speed of the wind driven generator;

the initial flow field data includes physical quantities on each computational grid;

the physical quantity comprises at least speed or/and pressure or/and temperature.

4. A wind driven generator simulation acceleration method based on reduced order decomposition processing according to claim 3, wherein,

in the second step, the method for constructing the initial data matrix is as follows:

step 21, obtaining physical quantity on each calculation grid;

step 22, the physical quantity which is taken out is stored into a digital matrix according to the requirement to form an initial data matrix;

the first column of the initial data matrix is the physical quantity of the 1 st time step, and the last column is the physical quantity of the N-1 st time step;

first of initial data matrixiThe row represents the time-dependent process information of the physical quantity on the ith computational grid, the first of whichjThe column represents the physical quantity of the jth time step.

5. The wind driven generator simulation acceleration method based on reduced order decomposition processing according to claim 1, wherein,

in the third step, the method for performing singular value decomposition on the initial data matrix is as follows:

step 31, performing singular value decomposition operation on the initial data matrix to obtain a singular value matrix;

step 32, setting singular values arranged from large to small on diagonal lines of the singular value matrix;

step 33, obtaining the energy contribution degree of each singular value to the initial data matrix;

and step 32, accumulating the energy contribution degrees one by one from large to small to obtain energy contribution sums, and when the energy contribution sums are larger than an energy threshold value, counting the number of the energy contribution degrees participating in accumulation, and taking the number of the energy contribution degrees as a reduction rank.

6. The wind driven generator simulation acceleration method based on reduced order decomposition processing according to claim 1, wherein,

the fourth step, the method for carrying out dynamic mode reduced order decomposition on the initial data matrix is as follows:

step 41, constructing an initial prediction matrix differing by one time step according to the initial data matrix;

step 42, according to the reduced rank, carrying out truncated singular value decomposition on the initial flow field data to obtain a decomposition result;

step 43, constructing a decomposition similarity matrix between the initial flow field data and the initial prediction matrix according to the decomposition result and the initial prediction matrix;

step 44, performing eigenvalue decomposition on the decomposition similarity matrix to obtain eigenvalues and eigenvectors;

and 45, constructing space-time characteristic quantity for representing the flow process of the wind driven generator according to the decomposition result, the initial prediction matrix, the characteristic value and the characteristic vector.

7. The wind driven generator simulation acceleration method based on reduced order decomposition processing according to claim 6, wherein,

the method for constructing the initial prediction matrix is as follows:

the first column of the initial prediction matrix is the flow field data of the 2 nd time step, and the last column is the flow field data of the last time step;

first of initial prediction matrixiThe row represents the time-dependent process information of the physical quantity on the ith computational grid, the first of whichjThe column represents the physical quantity of the j+1th time step.

8. The wind driven generator simulation acceleration method based on reduced order decomposition processing according to claim 7, wherein,

in the fifth step, the method for obtaining the predicted flow field data of the time step to be calculated is as follows:

step 51, constructing a reduced-order prediction matrix according to the space-time characteristic quantity, the initial flow field data and the initial prediction matrix, wherein the reduced-order prediction matrix comprises flow field data of a time step to be calculated;

step 52, calculating the reduced order prediction matrix to obtain predicted flow field data of the time step to be calculated;

step 53, adding the predicted flow field data of the time step to be calculated to the initial prediction matrix and the initial flow field data, and constructing a new initial prediction matrix and a new initial data matrix;

and step 54, obtaining predicted flow field data which is different from the new initial prediction matrix by one time step according to the new initial prediction matrix and the new initial data matrix, and circularly reciprocating to finish the calculation of a plurality of time step flow field data to be calculated.

9. A wind driven generator simulation acceleration method based on reduced order decomposition processing is characterized in that,

the method comprises the following steps:

carrying out full-order hydrodynamic simulation calculation on the wind driven generator by utilizing a pre-constructed hydrodynamic simulation model to obtain full-order initial flow field data of a certain time step;

processing initial flow field data through a pre-constructed reduced order calculation model, and constructing an initial data matrix; singular value decomposition is carried out on the initial data matrix to obtain a plurality of singular values, and a reduced rank is determined according to the energy duty ratio of the singular values;

predicting a reduced order model by utilizing pre-constructed data, and carrying out dynamic modal reduced order decomposition on an initial data matrix according to a reduced order rank to obtain space-time characteristic quantity for representing a flow process of a flow field; generating a reduced order prediction matrix according to the space-time characteristic quantity, and obtaining predicted flow field data of the time step to be calculated according to the reduced order prediction matrix;

and obtaining predicted flow field data through the reduced order calculation model and the data prediction reduced order model, and performing simulation calculation on the wind driven generator to realize simulation acceleration of the wind driven generator.

10. A wind driven generator simulation acceleration system based on reduced order decomposition processing is characterized in that,

comprising the following steps:

one or more processors;

a storage means for storing one or more programs;

the one or more programs, when executed by the one or more processors, cause the one or more processors to implement a reduced order decomposition process based wind turbine simulation acceleration method as claimed in any one of claims 1-9.

Images