A method for rapid simulation of temperature field reduction of GIS equipment

Abstract

The invention discloses a method and device for rapid simulation of temperature field reduction of GIS equipment. The method comprises the following steps: step S1, establishing a three-dimensional geometric simulation model of the object to be investigated according to its physical structure, and using finite element simulation software to perform multiple groups of working condition simulations on the established three-dimensional simulation model to obtain multiple groups of three-dimensional temperature field simulation analysis results; step S2, extracting the three-dimensional temperature field simulation results as a one-dimensional temperature field data group by feature point extraction and associating temperature information data on the feature points according to the established three-dimensional geometric simulation model and the three-dimensional temperature field simulation results; step S3, changing different input conditions to simulate through step S1, and constructing a training sample library according to the corresponding input one-dimensional data group and the one-dimensional temperature field data group extracted in step S2, constructing a training model, and using the samples of the training sample library to learn using machine learning software to generate a reduced-order model.

Description

Technical Field

[0001] The invention relates to the technical field of electrical engineering and electric data processing, and in particular to a method for rapid simulation of temperature field reduction of GIS (GASinsulated SWITCHGEAR, gas insulated fully enclosed switchgear) equipment. Background Art

[0002] Finite element simulation, as a general analysis method, can achieve high simulation model accuracy. However, for large models running three-dimensional transient analysis, this method is computationally expensive and consumes huge resources and time. Therefore, fast simulation technology is particularly important in three-dimensional transient real-time analysis.

[0003] Model reduction technology has been applied in the field of automatic control and circuit systems for a long time, and has always been one of the theoretical foundations of large-scale circuit design automation software. However, this basic and simple idea, as a kind of mathematical method with a relatively systematic theoretical basis, has only been around in recent years. How to transform a large-scale complex system into a smaller-scale approximate reduced-order system under certain conditions, and satisfy the requirement that the error between the reduced-order system and the original system is small enough, and keep the main performance of the original system such as stability, passivity and structural characteristics as much as possible, and at the same time, the reduction algorithm is stable and efficient, etc., is also a problem that the industry is currently committed to solving. Summary of the invention

[0004] In order to overcome the deficiencies of the above-mentioned prior art, the purpose of the present invention is to provide a method for reducing the order of rapid simulation of the temperature field of GIS equipment, so as to achieve the purpose of real-time simulation and analysis of the internal temperature distribution of GIS based on external temperature sensor data.

[0005] To achieve the above object, the present invention proposes a method for rapid simulation of temperature field reduction of GIS equipment, comprising the following steps:

[0006] Step S1, establishing a three-dimensional geometric simulation model of the object to be investigated according to its physical structure, and using finite element simulation software to perform multiple groups of working condition simulations on the established three-dimensional simulation model to obtain multiple groups of three-dimensional temperature field simulation analysis results;

[0007] Step S2, according to the established three-dimensional geometric simulation model and the three-dimensional simulation results of the temperature field, by extracting characteristic points and associating the temperature information data on the characteristic points, the three-dimensional simulation results of the temperature field are extracted into a one-dimensional data group of the temperature field, that is, a reduced-order correspondence relationship between the three-dimensional temperature field and the one-dimensional data group M of the temperature field is established;

[0008] Step S3, changing different input conditions to simulate through step S1, and constructing a training sample library according to the corresponding input one-dimensional data group and temperature field one-dimensional data group extracted according to step S2, constructing a training model and using the samples in the training sample library to learn using machine learning software to generate a reduced-order model.

[0009] Preferably, after step S3, the following steps are also included:

[0010] Step S4, selecting a group of data within the input range, obtaining an output data group through a reduced-order model, and assigning the output result to the three-dimensional geometric simulation model established in step S1 according to the extracted feature points, and inverting the three-dimensional temperature field distribution on the three-dimensional geometric simulation model according to the continuity of the temperature distribution;

[0011] Step S5, select a group of data within the input range, compare the inverted three-dimensional temperature field distribution obtained according to step S4 and the three-dimensional temperature field distribution of the finite element simulation obtained according to step S1, obtain the error of the temperature field distribution of the two, and determine whether to retrain to obtain a new reduced-order model based on the error.

[0012] Preferably, step S1 further comprises:

[0013] Step S100, establishing a simulation model for finite element / finite volume method temperature field simulation of the GIS device according to the physical structure of the GIS device under investigation;

[0014] Step S101, determine the input and output of the simulation model, run multiple groups of working conditions in the finite element simulation software to perform temperature field simulation on the simulation model using the finite element / finite volume method, and output the internal temperature field distribution of the GIS equipment.

[0015] Preferably, step S2 further comprises:

[0016] Step S200, splitting the model according to the complexity of the actual model based on the three-dimensional geometric simulation model established in step S1;

[0017] Step S201, according to the geometric features of the model, a cylindrical coordinate system is constructed with the center line of the internal conductor as the axis, and feature points are taken on the internal conductor, the shell, the insulator surface and the air chamber module. The points are determined by (l, r, θ), where l is the axial position in the cylindrical coordinates, r is the distance from the projection on the reference plane to the coordinate origin, and θ is the azimuth of the projection point;

[0018] Step S202, according to the position of the feature point and the three-dimensional simulation analysis result in step S1, the feature point is used as an intermediate connection quantity, the three-dimensional simulation analysis result is extracted into a one-dimensional data group, and a reduced-order correspondence relationship between the three-dimensional temperature field of the three-dimensional simulation analysis result and the one-dimensional data group is established;

[0019] Step S203, convert the input of the simulation model into a one-dimensional data set I = (u, i, e, T), where u is the effective value of the GIS three-phase voltage, i is the effective value of the GIS three-phase current, e is the external environment temperature, T = (t 1 ……t k4 ) is the measured value of the GIS casing temperature sensor.

[0020] Preferably, step S3 further comprises:

[0021] Step S300, changing different input conditions to perform simulation, obtaining the corresponding temperature field distribution calculation results through step S1, and extracting the corresponding input one-dimensional data group I and temperature field one-dimensional data group M according to step S2, and the group (I, M) constitutes a set of training library data;

[0022] Step S301, using the training sample library, through machine learning software, the training samples of the training sample library are run for multiple times to learn, establish the corresponding relationship between the input one-dimensional data group I and the temperature field one-dimensional data group M, reduce the three-dimensional simulation to a one-dimensional model that can be used for system simulation, and obtain the reduced-order model of the temperature field simulation.

[0023] Preferably, in step S4, a group of data within the input range is selected, and an output data group is obtained through the reduced-order model. The output result is assigned to the three-dimensional model established in step S1 according to the extracted feature points. According to the continuity of the temperature distribution, the three-dimensional temperature field distribution is inverted on the model using linear interpolation.

[0024] Preferably, the output result includes coordinate information of each feature point and corresponding temperature information.

[0025] Preferably, in step S5, a group of data within the input range is selected, and the inverted three-dimensional temperature field distribution is obtained according to step S4, and the three-dimensional temperature field distribution of finite element simulation is obtained according to step S1, and then the simulation results of the two are compared to obtain the errors of the temperature field distributions of the two; the maximum value and the average value of the error distribution are compared with the set error range. If the error exceeds the set range, the group of data is repeated in step S2 to extract the data, and it is added to the training data group in step S3, and the training is re-performed to obtain a new reduced-order model, and verification is performed again until the verification error is within the range.

[0026] In order to achieve the above object, the present invention also provides a device for rapid simulation of temperature field reduction of GIS equipment, comprising:

[0027] The model building simulation unit is used to build a three-dimensional geometric simulation model of the object of investigation according to the physical structure of the object of investigation, and use finite element simulation software to perform multiple groups of working condition simulations on the established three-dimensional simulation model to obtain multiple groups of three-dimensional temperature field simulation analysis results;

[0028] A reduction order processing unit is used to extract the three-dimensional simulation results of the temperature field into a one-dimensional data group of the temperature field by extracting characteristic points and associating the temperature information data on the characteristic points according to the established three-dimensional geometric simulation model and the three-dimensional simulation results of the temperature field, that is, to establish a reduced order correspondence between the three-dimensional temperature field and the one-dimensional data group M of the temperature field;

[0029] The reduced-order model generation unit is used to change different input conditions to simulate through the model construction simulation unit, and to construct a training sample library based on the corresponding input one-dimensional data group and temperature field one-dimensional data group extracted by the reduced-order processing unit, to build a training model and use the samples in the training sample library to learn using machine learning software to generate a reduced-order model.

[0030] Preferably, the device further comprises:

[0031] An inversion unit is used to select a group of data within an input range, obtain an output data group through a reduced-order model, and assign the output result to a three-dimensional geometric simulation model established by the model building simulation unit according to the extracted characteristic points, and invert the three-dimensional temperature field distribution on the three-dimensional geometric simulation model according to the continuity of the temperature distribution;

[0032] The accuracy verification unit is used to select a set of data within the input range, compare the three-dimensional temperature field distribution obtained by the inversion unit and the three-dimensional temperature field distribution obtained by the finite element simulation according to the model construction simulation unit, obtain the error of the temperature field distribution of the two, and determine whether to retrain to obtain a new reduced-order model based on the error.

[0033] Compared with the prior art, the method and device for rapid simulation of temperature field reduction of GIS equipment of the present invention are based on the concept of order reduction, and achieve the purpose of real-time simulation and analysis of temperature distribution inside GIS based on external temperature sensor data. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 A flowchart of the steps of a method for rapid simulation of temperature field reduction of GIS equipment according to the present invention;

[0035] Figure 2 A schematic diagram of a three-dimensional geometric model constructed for a specific embodiment of the present invention;

[0036] Figure 3 It is a schematic diagram of the simulation structure of a specific embodiment of the present invention in finite element simulation software;

[0037] Figure 4 A schematic diagram of accuracy verification in a specific embodiment of the present invention;

[0038] Figure 5 This is a system structure diagram of a device for rapid simulation of temperature field reduction of GIS equipment according to the present invention. DETAILED DESCRIPTION

[0039] The following describes the implementation of the present invention through specific examples and in conjunction with the accompanying drawings. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific examples, and the details in this specification can also be modified and changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention.

[0040] Figure 1 This is a flowchart of the steps of a method for rapid simulation of temperature field reduction of GIS equipment according to the present invention. Figure 1 As shown, the present invention provides a method for rapid simulation of temperature field reduction of GIS equipment, comprising the following steps:

[0041] Step S1, based on the physical structure of the object under investigation, a three-dimensional geometric simulation model of the object is established, and finite element simulation software is used to perform multiple groups of working condition simulations on the established three-dimensional geometric simulation model to obtain multiple groups of three-dimensional temperature field simulation analysis results, so as to serve as training samples for the reduced-order model of step S3.

[0042] Specifically, step S1 further includes:

[0043] Step S100, based on the physical structure of the GIS device under investigation, a three-dimensional geometric simulation model for finite element / finite volume method temperature field simulation of the GIS device is established.

[0044] In the present invention, the object of investigation is GIS equipment, and a three-dimensional geometric simulation model of the GIS equipment is established according to the physical structure of the GIS equipment. In the present invention, since GIS heating failure usually occurs at the busbar joint, a three-dimensional geometric simulation model of the GIS equipment is constructed. Figure 2 The three-dimensional geometric simulation model shown is constructed using existing known technologies and will not be described in detail here.

[0045] Step S101, determine the input and output of the simulation model, perform temperature field simulation on the simulation model using the finite element / finite volume method, and output the internal temperature field distribution of the GIS equipment.

[0046] In the present invention, the input of the simulation model is GIS voltage, current, ambient temperature and shell temperature, and the output is the temperature field distribution inside the GIS. Generally speaking, the temperature field simulation requires setting boundary conditions. The external ambient temperature is the constraint boundary condition of the entire simulation model, which is called the external boundary condition here. That is to say, the outer boundary of the temperature field calculation of the GIS equipment is the external ambient temperature, and its input is the temperature at different positions of the shell. The simulation result is the temperature distribution of the internal conductor and the air chamber.

[0047] In a specific embodiment of the present invention, the distribution of current, voltage, ambient temperature and shell temperature can be monitored by sensors (each sensor is set on the GIS physical device, the position of each sensor is correspondingly set on the model, and the data monitored by each sensor is uploaded to the corresponding position of the model).

[0048] Run multiple groups of working conditions in the finite element simulation software (multiple groups of working conditions refer to multiple groups of data including current, voltage, ambient temperature, and shell temperature). The range of volume power (volume power is obtained by current and voltage processing), the range of external ambient temperature, and the range of shell temperature distribution are determined artificially, covering the maximum and minimum values ​​of the actual situation (that is, determining the upper and lower limits of each input value). That is, the boundary condition values ​​of the simulation analysis are determined according to the actual operation conditions, and the temperature field simulation is performed using the finite element / finite volume method. After the calculation is completed, each group of calculation result files is obtained, and the results of the three-dimensional temperature field simulation analysis of each group can be obtained, such as Figure 3 The data bin file of the simulation results shown is not elaborated here because it is the prior art to simulate each group of working conditions by finite element simulation software through finite element / finite volume method to obtain the three-dimensional temperature field distribution.

[0049] Step S2, based on the established three-dimensional geometric simulation model and the three-dimensional simulation results of the temperature field, the three-dimensional simulation results of the temperature field are extracted into a one-dimensional data group of the temperature field by extracting feature points and associating the temperature information data on the feature points, that is, a reduced-order correspondence between the three-dimensional temperature field and the one-dimensional data group M of the temperature field is established.

[0050] Specifically, step S2 further includes:

[0051] Step S200, based on the three-dimensional simulation model established in step S1, split the model according to the complexity of the actual model.

[0052] In the present invention, the GIS equipment model is split into a shell 1, ...N1, a conductor 1, ...N2, an insulator 1, ..., N3, and an air chamber 1, ...N4 according to the complexity of the actual model. Specifically, the internal conductor of the model is divided into four sections D1, D2, D3, and D4 according to the joints; the corresponding shell is also divided into four sections L1, L2, L3, and L4; insulators S1, S2, S3, and S4; and air chambers N1, N2, N3, and N4. Of course, the specific splitting can be carried out according to the actual situation of the model, and the present invention is not limited to this.

[0053] Step S201, according to the geometric features of the model, a cylindrical coordinate system is constructed with the center line of the internal conductor as the axis, and feature points are taken on the internal conductor, the outer shell, the insulator surface and the air chamber module. The points are determined by (l, r, θ), where l is the axial position in the cylindrical coordinates (the distance from the reference plane), r is the distance from the projection on the reference plane to the coordinate origin, and θ is the azimuth of the projection point.

[0054] Step S202, based on the position of the feature point and the three-dimensional simulation analysis results in step S1, the feature point is used as an intermediate link quantity, the three-dimensional simulation analysis results are extracted into a one-dimensional array, and a reduced-order correspondence between the three-dimensional temperature field and the one-dimensional array m = (X, Y, Z) is established, that is, m is the temperature data information corresponding to the current feature point, and the three-dimensional temperature distribution is determined by the feature point set and the temperature information set m on the feature point.

[0055] That is to say, when the GIS equipment model is split into shell 1, ... N1, conductor 1, ... N2, insulator 1, ..., N3, air chamber 1, ... N4, considering the continuity of temperature distribution and the symmetry of GIS structure design, each module can be represented by a corresponding one-dimensional data set m i (X,Y,Z)(where X=(x……x k ),Y=(y……y k ),Z=(z……z k ), k represents the number of characteristic points on the i-th module) corresponds to the temperature distribution on the characterization model. Through the above steps, the three-dimensional temperature field is extracted into a one-dimensional temperature field data set M = (m1, ...m n ).

[0056] Figure 2 Taking the model as an example, first, according to the structure of the three-dimensional geometric model, feature points are taken on each module of the model (the more feature points, the higher the accuracy of characterizing the three-dimensional structure, and the number of feature points is selected according to the actual situation). The position information of each feature point is determined by (l, r, θ). The position information of all feature points constitutes a one-dimensional point set, and the simulation information temperature value, temperature change gradient and change direction (x, y, z) of the corresponding points are associated at the same time. The parameters of all points constitute a one-dimensional parameter set, and the three-dimensional simulation results are converted into a one-dimensional point set and a one-dimensional parameter set.

[0057] Step S203, convert the input of the simulation model into a one-dimensional data set I = (u, i, e, T), where u is the effective value of the GIS three-phase voltage, i is the effective value of the GIS three-phase current, e is the external environment temperature, T = (t 1 ……t k4 ) is the measured value of the GIS casing temperature sensor.

[0058] Step S3, changing different input conditions to simulate through step S1, and extracting the corresponding input one-dimensional data group I and temperature field one-dimensional data group M according to step S2 to construct a training sample library; constructing a training model, and using the training samples in the training sample library to learn using machine learning software to generate a reduced-order model.

[0059] Specifically, step S3 further includes:

[0060] Step S300, changing different input conditions (current, voltage, ambient temperature, shell temperature distribution) to perform simulation, obtaining the corresponding temperature field distribution calculation results through step S1, and extracting the corresponding input one-dimensional data group I and temperature field one-dimensional data group M according to step S2, and the group (I, M) constitutes a set of training library data. In the present invention, the data of the training library includes normal data and fault data, the normal data refers to the uniform distribution of the shell temperature, and the fault data refers to the uneven distribution of the shell temperature.

[0061] Under the same cylindrical coordinates, take the characteristic points different from those in step S2 on the shell to establish the input data set I (current, voltage, ambient temperature, shell temperature distribution) = (u, i, e, T), where u is the effective value of the GIS three-phase voltage, i is the effective value of the GIS three-phase current, e is the external ambient temperature, T = (t 1 ……t k4 ) is the measured value of the temperature sensor at the corresponding characteristic point of the shell.

[0062] In step S300, constructing the training library also includes determining the upper and lower limits of each input value in combination with the operating conditions of the actual GIS equipment, which serve as the boundaries of the reduced-order model.

[0063] Step S301, using the training sample library, through machine learning software, the training library data is run for multiple times to learn, establish the corresponding relationship between the input one-dimensional data group I and the temperature field one-dimensional data group M, reduce the three-dimensional simulation to a one-dimensional model that can be used for system simulation, and obtain a reduced-order model for temperature field simulation. In a specific embodiment of the present invention, the machine learning technology includes deep learning, SVD, LTI, and recursive neural network technology.

[0064] Specifically, several groups of different working conditions are set in the finite element simulation software (i.e., the input data group I (current, voltage, ambient temperature, shell temperature distribution) = (u, i, e, T)). In the specific embodiment of the present invention, 9 groups of working conditions are set (of course, more can be set. The more training data, the higher the accuracy of the reduced-order model). Simulation calculations are performed separately to obtain temperature field distribution simulation results, and then the results M are one-dimensionalized through step S2 to form 9 groups of training samples (I, M).

[0065] In the present invention, the boundary of the simulation result is determined according to the maximum and minimum values ​​of the actual situation (that is, the upper and lower limits of each input value are determined), which is used as the boundary of the reduced-order model.

[0066] Through machine learning software, learning is performed based on the 9 sets of training samples obtained, and when learning is completed, a reduced-order model is generated.

[0067] For example: assuming that 9 working conditions are simulated in step S1, and the degradation model is obtained by training based on these 9 sets of data, then for the 10th working condition, the temperature feature set of the characteristic points can be directly obtained based on the trained reduced-order model, and then the three-dimensional temperature distribution can be obtained by inverting the characteristic point set. It is no longer necessary to perform the three-dimensional temperature field calculation in step S1, which greatly improves the simulation speed. The more training data there is, the more accurate the final simulation result will be.

[0068] Step S3 is to train the one-dimensional result of the simulation result of step S1 in step S2 through machine learning. Since there are countless working conditions in actual situations (voltage, current, and temperature are all continuous values), in this embodiment, 9 groups of typical working conditions are selected as training sets and test sets for machine learning (the machine learning algorithm adopts a mature calculation method of python, which will not be described in detail here). After obtaining the training model, the working condition can be determined and the corresponding temperature distribution can be obtained by the model without the need to perform temperature field simulation. The reason for this is that the simulation analysis takes a long time to calculate and requires a large amount of computing resources, while the temperature field is obtained quickly through the training model and occupies less resources, which is also the main purpose of performing step S2.

[0069] Since machine learning utilizes existing mature technologies, the present invention will not elaborate on this.

[0070] Step S4, select a group of data within the input range, obtain an output data group through a reduced-order model, and assign the output result to the three-dimensional model established in step S1 according to the extracted feature points, and invert the three-dimensional temperature field distribution on the three-dimensional model according to the continuity of the temperature distribution.

[0071] In the present invention, a set of data within the input range (i.e. (current, voltage, ambient temperature, shell temperature distribution) = (u, i, e, T)) is selected, and an output data set is obtained through the reduced-order model. The output result file is in .bin format binary, which contains the coordinate information (l, r, θ) and corresponding temperature information (x, y, z) of each feature point.

[0072] The output result is assigned to the three-dimensional geometric simulation model established in step S1 according to the extracted feature points, and the three-dimensional temperature field distribution is inverted on the model by linear interpolation according to the continuity of the temperature distribution.

[0073] Step S5, select a group of data within the input range, compare the inverted three-dimensional temperature field distribution obtained according to step S4 and the three-dimensional temperature field distribution of the finite element simulation obtained according to step S1, obtain the error of the temperature field distribution of the two, and determine whether to retrain to obtain a new reduced-order model based on the error.

[0074] Specifically, a set of data within the input range is selected, and the inverted three-dimensional temperature field distribution is obtained according to step S4, and the three-dimensional temperature field distribution of the finite element simulation is obtained according to step S1, and then the simulation results of the two are compared to obtain the error of the temperature field distribution of the two. Figure 4 shown.

[0075] Compare the maximum and average values ​​of the error distribution with the set error range. If the error exceeds the set range, repeat step S2 to extract data for this group of data, add it to the training data group in step S3, retrain to obtain a new reduced-order model, and verify again until the verification error is within the range.

[0076] Figure 5 This is a system structure diagram of a device for rapid simulation of temperature field reduction of GIS equipment according to the present invention. Figure 5 As shown, the present invention provides a device for rapid simulation of temperature field reduction of GIS equipment, comprising:

[0077] The model building simulation unit 501 is used to establish a three-dimensional simulation model of the object according to the physical structure of the object under investigation, and use finite element simulation software to perform multiple groups of working condition simulations on the established three-dimensional simulation model to obtain multiple groups of three-dimensional temperature field simulation analysis results, so as to serve as training samples for the reduced-order model.

[0078] Specifically, the model building simulation unit 501 further includes:

[0079] The simulation model building module is used to build a simulation model for finite element / finite volume method temperature field simulation of GIS equipment according to the physical structure of the GIS equipment under investigation.

[0080] In the present invention, the object of investigation is GIS equipment. Based on the physical structure of the GIS equipment, a three-dimensional simulation model of the GIS equipment is established. Since GIS heating failure usually occurs at the busbar joint, in a specific embodiment of the present invention, a three-dimensional simulation model of the GIS equipment is constructed. Figure 2 The three-dimensional geometric simulation model shown is constructed using existing known technologies and will not be described in detail here.

[0081] The simulation module is used to determine the input and output of the simulation model, use the finite element / finite volume method to simulate the temperature field of the simulation model, and output the internal temperature field distribution of the GIS equipment.

[0082] In the present invention, the input of the simulation model is GIS voltage, current, ambient temperature and shell temperature, and the output is the temperature field distribution inside the GIS. Generally speaking, the temperature field simulation requires setting boundary conditions. The external ambient temperature is the constraint boundary condition of the entire simulation model, which is called the external boundary condition here. That is to say, the outer boundary of the temperature field calculation of the GIS equipment is the external ambient temperature, and its input is the temperature at different positions of the shell. The simulation result is the temperature distribution of the internal conductor and the air chamber.

[0083] In a specific embodiment of the present invention, the distribution of current, voltage, ambient temperature, and shell temperature can all be monitored by sensors (each sensor is set on a GIS physical device, the position of each sensor is correspondingly set on the model, and the data monitored by each sensor is uploaded to the corresponding position of the model) as preparation for reduced-order data and input of the reduced-order training model.

[0084] Run multiple groups of working conditions in the finite element simulation software (multiple groups of working conditions refer to multiple groups of data including current, voltage, ambient temperature, and shell temperature). The range of volume power (volume power is obtained by current and voltage processing calculation), the range of external ambient temperature, and the range of shell temperature distribution can be determined manually, covering the maximum and minimum values ​​of the actual situation (that is, determining the upper and lower limits of each input value). That is, the boundary condition values ​​of the simulation analysis are determined according to the actual operation conditions, and the temperature field simulation is performed using the finite element / finite volume method. After the calculation is completed, each group of calculation result files is obtained, that is, the results of the three-dimensional temperature field simulation analysis of each group are obtained, such as Figure 3 The data bin file of the simulation results shown is not elaborated here because it is the prior art to simulate each group of working conditions by finite element simulation software through finite element / finite volume method to obtain the three-dimensional temperature field distribution.

[0085] The reduced-order processing unit 502 is used to extract the three-dimensional simulation results of the temperature field into a one-dimensional data group of the temperature field based on the established three-dimensional simulation model and the three-dimensional simulation results of the temperature field of the model construction simulation unit 501, through feature point extraction and association of temperature information data on the feature points, that is, to establish a reduced-order correspondence between the three-dimensional temperature field and the one-dimensional data group M of the temperature field.

[0086] The order reduction processing unit 502 is specifically used for:

[0087] According to the three-dimensional geometric simulation model established by the model building simulation unit 501, the model is split according to the complexity of the actual model.

[0088] In the present invention, the GIS equipment model is split into a shell 1, ...N1, a conductor 1, ...N2, an insulator 1, ..., N3, and an air chamber 1, ...N4 according to the complexity of the actual model. Specifically, the internal conductor of the model is divided into four sections D1, D2, D3, and D4 according to the joints; the corresponding shell is also divided into four sections L1, L2, L3, and L4; insulators S1, S2, S3, and S4; and air chambers N1, N2, N3, and N4.

[0089] According to the geometric characteristics of the model, a cylindrical coordinate system is constructed with the center line of the internal conductor as the axis. Feature points are taken on the internal conductor, shell, insulator surface and air chamber module. The points are determined by (l, r, θ), where l is the axial position in the cylindrical coordinates (the distance from the reference plane), r is the distance from the projection on the reference plane to the coordinate origin, and θ is the azimuth of the projection point.

[0090] The three-dimensional simulation analysis results in the simulation unit 501 are constructed according to the positions of the feature points and the model. The feature points are used as intermediate connecting quantities to extract the three-dimensional simulation analysis results into a one-dimensional array, and a reduced-order correspondence between the three-dimensional temperature field and the one-dimensional array m = (X, Y, Z) is established, that is, m is the temperature data information corresponding to the feature point, and the three-dimensional temperature distribution is determined by the feature point set and the temperature information set m on the feature point.

[0091] That is to say, when the GIS equipment model is split into shell 1, ... N1, conductor 1, ... N2, insulator 1, ..., N3, air chamber 1, ... N4, considering the continuity of temperature distribution and the symmetry of GIS structure design, each module can be represented by a corresponding one-dimensional data set m i (X,Y,Z)(where X=(x……x k ),Y=(y……y k ),Z=(z……z k ), k represents the number of characteristic points on the i-th module) corresponds to the temperature distribution on the characterization model. Through the above steps, the three-dimensional temperature field is extracted into a one-dimensional temperature field data set M = (m1, ...m n ).

[0092] Figure 2Taking the model as an example, first, according to the structure of the three-dimensional geometric model, feature points are taken on each module of the model (the more feature points, the higher the accuracy of characterizing the three-dimensional structure, and the number of feature points is selected according to the actual situation). The position information of each feature point is determined by (l, r, θ). The position information of all feature points constitutes a one-dimensional point set, and the simulation information temperature value, temperature change gradient and change direction (x, y, z) of the corresponding point are associated. The parameters of all points constitute a one-dimensional parameter set, and the three-dimensional simulation results are converted into a one-dimensional point set and a point parameter set, that is, the above-mentioned M.

[0093] The input of the simulation model is converted into a one-dimensional data set I = (u, i, e, T), where u is the effective value of the GIS three-phase voltage, i is the effective value of the GIS three-phase current, e is the external environment temperature, and T = (t 1 ……t k4 ) is the measured value of the GIS casing temperature sensor.

[0094] The reduced-order model generating unit 503 is used to change different input conditions to simulate through the model building simulation unit 501, and construct a training sample library according to the corresponding input one-dimensional data group I and temperature field one-dimensional data group M extracted by the reduced-order processing unit, and construct a training model, and use the samples in the training sample library to learn using machine learning software to generate a reduced-order model.

[0095] Specifically, the reduced-order model generating unit 503 is specifically used for:

[0096] By changing different input conditions for simulation, the corresponding temperature field distribution calculation results are obtained through the model construction simulation unit, and the corresponding input one-dimensional data group I and the temperature field one-dimensional data group M are extracted according to the order reduction processing unit, and the group (I, M) constitutes a set of training library data. In the present invention, the data of the training library includes normal data and fault data, the normal data refers to the uniform distribution of the shell temperature, and the fault data refers to the uneven distribution of the shell temperature.

[0097] Specifically, under the same cylindrical coordinates, different characteristic points in the reduced-order processing unit are taken on the shell to establish the input data set I (current, voltage, ambient temperature, shell temperature distribution) = (u, i, e, T), where u is the effective value of the GIS three-phase voltage, i is the effective value of the GIS three-phase current, e is the external ambient temperature, T = (t 1 ……t k4 ) is the measured value of the temperature sensor at the corresponding characteristic point of the shell.

[0098] Building a training library also includes determining the upper and lower limits of the input values ​​based on the actual operating conditions of the GIS equipment, which serve as the boundaries of the reduced-order model.

[0099] By using the training sample library and machine learning software, the training library data is run multiple times for learning, the corresponding relationship between the input one-dimensional data group I and the temperature field one-dimensional data group M is established, the three-dimensional simulation is reduced to a one-dimensional model that can be used for system simulation, and the reduced-order model of the temperature field simulation is obtained. In a specific embodiment of the present invention, the machine learning technology includes deep learning, SVD, LTI, and recursive neural network technology.

[0100] Specifically, several groups of different working conditions are set in the finite element simulation software (i.e., the input data group I (current, voltage, ambient temperature, shell temperature distribution) = (u, i, e, T)). In the specific embodiment of the present invention, 9 groups of working conditions are set (of course, more can be set. The more training data, the higher the accuracy of the reduced-order model). Simulation calculations are performed separately to obtain temperature field distribution simulation results, and then 9 groups of training samples (I, M) are formed through the one-dimensional result M of the reduced-order processing unit.

[0101] In the present invention, the boundary of the simulation result is determined according to the maximum and minimum values ​​of the actual situation (that is, the upper and lower limits of each input value are determined), which is used as the boundary of the reduced-order model.

[0102] Through machine learning software, learning is performed based on the 9 sets of training samples obtained, and when learning is completed, a reduced-order model is generated.

[0103] For example: assuming that 9 working conditions are simulated in the model building simulation unit, and a reduced-order model is obtained by training based on these 9 sets of data, then for the 10th working condition, the temperature feature set of the characteristic points can be directly obtained based on the trained reduced-order model, and then the three-dimensional temperature distribution can be obtained by inverting the characteristic point set. It is no longer necessary to perform the three-dimensional temperature field calculation in step S1, which greatly improves the simulation speed. The more training data there is, the more accurate the final simulation result will be.

[0104] The inversion unit 504 is used to select a group of data within the input range, obtain an output data group through a reduced-order model, and assign the output results to the three-dimensional geometric simulation model established by the model building simulation unit 501 according to the extracted feature points, and invert the three-dimensional temperature field distribution on the three-dimensional geometric simulation model according to the continuity of the temperature distribution.

[0105] In the present invention, a group of data within the input range is selected, and an output data group is obtained through the reduced-order model. The output result file is a binary file in .bin format, which contains the coordinate information (l, r, θ) and corresponding temperature information (x, y, z) of each feature point.

[0106] The output result is assigned to the three-dimensional geometric simulation model established in the model building simulation unit 501, and according to the continuity of the temperature distribution, the three-dimensional temperature field distribution is inverted on the model by using linear interpolation.

[0107] The accuracy verification unit 505 is used to select a group of data within the input range, compare the three-dimensional temperature field distribution obtained by the inversion unit 504 and the three-dimensional temperature field distribution obtained by the finite element simulation according to the model construction simulation unit 501, obtain the error of the temperature field distribution of the two, and determine whether to retrain to obtain a new reduced-order model based on the error.

[0108] Specifically, a set of data within the input range is selected, and the inverted three-dimensional temperature field distribution is obtained according to the inversion unit 504, and the three-dimensional temperature field distribution of the finite element simulation is obtained according to the model construction simulation unit 501, and then the simulation results of the two are compared to obtain the error of the temperature field distribution of the two. Figure 4 shown.

[0109] Compare the maximum and average values ​​of the error distribution with the set error range. If the error exceeds the set range, the data group is repeatedly extracted by the order reduction processing unit 502 and added to the training data group in the reduced-order model generation unit 503. Re-training is performed to obtain a new reduced-order model, which is verified again until the error is within the range.

[0110] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Any person skilled in the art may modify and alter the above embodiments without violating the spirit and scope of the present invention. Therefore, the scope of protection of the present invention shall be as set forth in the claims.

Claims

1. A method for rapid simulation of temperature field reduction of GIS equipment, comprising the following steps: Step S1, establishing a three-dimensional geometric simulation model of the object to be investigated according to its physical structure, and using finite element simulation software to perform multiple groups of working condition simulations on the established three-dimensional simulation model to obtain multiple groups of three-dimensional temperature field simulation analysis results; Step S2, according to the established three-dimensional geometric simulation model and the three-dimensional simulation results of the temperature field, by extracting characteristic points and associating the temperature information data on the characteristic points, extracting the three-dimensional simulation results of the temperature field into a one-dimensional data group of the temperature field, that is, establishing a reduced-order correspondence between the three-dimensional temperature field and the one-dimensional data group of the temperature field; Step S3, changing different input conditions to simulate step S1, and constructing a training sample library according to the corresponding input one-dimensional data group and temperature field one-dimensional data group extracted in step S2, constructing a training model, and using the samples in the training sample library to learn using machine learning software to generate a reduced-order model; Wherein, the step S2 further comprises: Step S200, based on the three-dimensional geometric simulation model established in step S1, split the model according to the complexity of the actual model; Step S201, according to the geometric features of the model, a cylindrical coordinate system is constructed with the center line of the internal conductor as the axis, and feature points are taken on the internal conductor, the shell, the insulator surface and the air chamber module. The points are determined by (l, r, θ), where l is the axial position in the cylindrical coordinates, r is the distance from the projection on the reference plane to the coordinate origin, and θ is the azimuth of the projection point; Step S202, according to the position of the feature point and the three-dimensional simulation analysis result in step S1, the feature point is used as an intermediate connection quantity, the three-dimensional simulation analysis result is extracted into a one-dimensional data group, and a reduced-order correspondence relationship between the three-dimensional temperature field of the three-dimensional simulation analysis result and the one-dimensional data group is established; Step S203, convert the input of the simulation model into a one-dimensional data set I = (u, i, e, T), where u is the effective value of the GIS three-phase voltage, i is the effective value of the GIS three-phase current, e is the external environment temperature, T = (t1...t k4 ) is the measured value of the GIS casing temperature sensor.

2. A method for rapid simulation of temperature field reduction of GIS equipment as claimed in claim 1, characterized in that: After step S3, the method further includes the following steps: Step S4, selecting a group of data within the input range, obtaining an output data group through a reduced-order model, and assigning the output result to the three-dimensional geometric simulation model established in step S1 according to the extracted feature points, and inverting the three-dimensional temperature field distribution on the three-dimensional geometric simulation model according to the continuity of the temperature distribution; Step S5, select a group of data within the input range, compare the inverted three-dimensional temperature field distribution obtained according to step S4 and the three-dimensional temperature field distribution of the finite element simulation obtained according to step S1, obtain the error of the temperature field distribution of the two, and determine whether to retrain to obtain a new reduced-order model based on the error.

3. A method for rapid simulation of temperature field reduction of GIS equipment as claimed in claim 2, characterized in that: Step S1 further comprises: Step S100, establishing a simulation model for finite element / finite volume method temperature field simulation of the GIS device according to the physical structure of the GIS device under investigation; Step S101, determine the input and output of the simulation model, run multiple groups of working conditions in the finite element simulation software to perform temperature field simulation on the simulation model using the finite element / finite volume method, and output the internal temperature field distribution of the GIS equipment.

4. A method for rapid simulation of temperature field reduction of GIS equipment as claimed in claim 1, characterized in that: Step S3 further comprises: Step S300, changing different input conditions to perform simulation, obtaining the corresponding temperature field distribution calculation results through step S1, and extracting the corresponding input one-dimensional data group I and temperature field one-dimensional data group M according to step S2, and the group (I, M) constitutes a set of training library data; Step S301, using the training sample library, through machine learning software, the training samples of the training sample library are run for multiple times to learn, establish the corresponding relationship between the input one-dimensional data group I and the temperature field one-dimensional data group M, reduce the three-dimensional simulation to a one-dimensional model that can be used for system simulation, and obtain the reduced-order model of the temperature field simulation.

5. A method for rapid simulation of temperature field reduction of GIS equipment as claimed in claim 4, characterized in that: In step S4, a group of data within the input range is selected, and an output data group is obtained through the reduced-order model. The output result is assigned to the three-dimensional model established in step S1 according to the extracted feature points. According to the continuity of the temperature distribution, the three-dimensional temperature field distribution is inverted on the model using linear interpolation.

6. A method for rapid simulation of temperature field reduction of GIS equipment as claimed in claim 4, characterized in that: The output result includes the coordinate information of each feature point and the corresponding temperature information.

7. A method for rapid simulation of temperature field reduction of GIS equipment as claimed in claim 4, characterized in that: In step S5, a group of data within the input range is selected, and the inverted three-dimensional temperature field distribution is obtained according to step S4, and the three-dimensional temperature field distribution of finite element simulation is obtained according to step S1, and then the simulation results of the two are compared to obtain the errors of the temperature field distributions of the two; the maximum value and the average value of the error distribution are compared with the set error range. If the error exceeds the set range, the group of data is repeated in step S2 to extract the data, and it is added to the training data group in step S3, and the training is re-performed to obtain a new reduced-order model, and verification is performed again until the verification error is within the range.

8. A device for rapid simulation of temperature field reduction of GIS equipment, comprising: The model building simulation unit is used to build a three-dimensional geometric simulation model of the object of investigation according to the physical structure of the object of investigation, and use finite element simulation software to perform multiple groups of working condition simulations on the established three-dimensional simulation model to obtain multiple groups of three-dimensional temperature field simulation analysis results; A reduction order processing unit is used to extract the three-dimensional simulation results of the temperature field into a one-dimensional data group of the temperature field by extracting characteristic points and associating the temperature information data on the characteristic points according to the established three-dimensional geometric simulation model and the three-dimensional simulation results of the temperature field, that is, to establish a reduced order correspondence between the three-dimensional temperature field and the one-dimensional data group M of the temperature field; A reduced-order model generation unit is used to change different input conditions to simulate through the model building simulation unit, and extract corresponding input one-dimensional data groups and temperature field one-dimensional data groups according to the reduced-order processing unit to build a training sample library, build a training model, and use the samples of the training sample library to learn using machine learning software to generate a reduced-order model; Wherein, the order reduction processing unit is used to: Based on the established 3D geometric simulation model, split the model according to the complexity of the actual model; According to the geometric characteristics of the model, a cylindrical coordinate system is constructed with the center line of the internal conductor as the axis. Feature points are taken on the internal conductor, shell, insulator surface and air chamber module. The points are determined by (l, r, θ), where l is the axial position in the cylindrical coordinates, r is the distance from the projection on the reference plane to the origin of the coordinates, and θ is the azimuth of the projection point. According to the position of the feature points and the three-dimensional simulation analysis results, the feature points are used as the intermediate connection quantity, the three-dimensional simulation analysis results are extracted into a one-dimensional data group, and the reduced-order correspondence between the three-dimensional temperature field of the three-dimensional simulation analysis results and the one-dimensional data group is established; The input of the simulation model is converted into a one-dimensional data set I = (u, i, e, T), where u is the effective value of the GIS three-phase voltage, i is the effective value of the GIS three-phase current, e is the external environment temperature, T = (t1... k4 ) is the measured value of the GIS casing temperature sensor.

9. The device for rapid simulation of temperature field reduction of GIS equipment as claimed in claim 8, characterized in that: The device also includes: An inversion unit is used to select a group of data within an input range, obtain an output data group through a reduced-order model, and assign the output result to a three-dimensional geometric simulation model established by the model building simulation unit according to the extracted characteristic points, and invert the three-dimensional temperature field distribution on the three-dimensional geometric simulation model according to the continuity of the temperature distribution; The accuracy verification unit is used to select a set of data within the input range, compare the three-dimensional temperature field distribution obtained by the inversion unit and the three-dimensional temperature field distribution obtained by the finite element simulation according to the model construction simulation unit, obtain the error of the temperature field distribution of the two, and determine whether to retrain to obtain a new reduced-order model based on the error.

Images