SF6 density relay temperature compensation method

Abstract

The invention discloses an SF6 density relay temperature compensation method based on finite element simulation and a GA-PSO-RBF algorithm, and belongs to the technical field of power equipment simulation and temperature compensation. Comprising the following steps: establishing a COMSOL air chamber model; optimizing RBF neural network parameters by using a genetic particle swarm; constructing a genetic particle swarm RBF neural network air chamber compensation model; outputting an air chamber internal temperature compensation value, wherein the neural network model is obtained by training based on a pre-constructed data set; the data set is obtained by simulating different SF6 gas chamber structure parameters and environment temperature changes through finite element simulation software COMSOL. According to the invention, the actual internal temperature of SF6 can be compensated simply, efficiently, quickly and accurately.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment simulation and temperature compensation, and in particular to a temperature compensation method for an SF6 density relay, which is used to optimize the prediction accuracy and dynamic compensation of the temperature field inside an SF6 gas chamber under a variable temperature environment. Background Art

[0002] In high-voltage electrical equipment, the density of SF6 gas is a key parameter affecting the insulation performance and arc-extinguishing capability of the gas chamber, and must be strictly monitored during equipment operation and maintenance. SF6 gas density is primarily monitored by installing SF6 density relays on GIS equipment. These relays convert the pressure within the gas chamber to an equivalent pressure at 20°C, thereby indicating the density of the SF6 gas. However, GIS equipment is a precision device and cannot incorporate a built-in temperature sensor. Due to the high specific heat capacity of SF6 gas, when the external temperature changes, the temperature response within the gas chamber lags behind the external temperature change, exhibiting a certain degree of hysteresis. This phenomenon affects the accuracy of density measurements, making it difficult to distinguish whether deviations in the SF6 density relay readings are caused by faults or other factors.

[0003] In recent years, few domestic and international researchers have studied the temperature hysteresis of SF6 density relays, primarily relying on the experience of testers to compensate for them. Key issues include: traditional compensation relies too heavily on tester experience, resulting in significant errors; and the varying temperature hysteresis effects of chambers with different structural parameters make it difficult to maintain consistent compensation results across different environments. Therefore, determining how to eliminate the impact of ambient temperature changes on the internal temperature of the SF6 chamber and correct temperature measurement accuracy is crucial for accurately assessing the operating status of SF6 density relays. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to solve the low precision and poor adaptability of the traditional SF6 density relay temperature compensation method and improve the reliability of the device operation.

[0005] In order to solve the technical problem, the solution of the present invention is to provide a temperature compensation method for SF6 density relay based on high-precision finite element simulation and neural network optimization.

[0006] The simulation of the internal temperature field distribution characteristics of the SF6 gas chamber in a variable temperature environment using a high-precision finite element simulation model includes the following steps:

[0007] (1) Geometric modeling of the SF6 gas chamber is performed in the multi-physics coupling simulation software; the material properties of the SF6 gas chamber are assigned, including thermal conductivity, specific heat capacity, density, etc. The material parameters are based on the physical properties of the SF6 gas and the gas chamber shell material; the external temperature is fitted using Fourier transform to obtain the external temperature change curve; in the solid and fluid heat transfer module, the outer wall temperature of the SF6 gas chamber is set to the external temperature change curve, and the initial temperature of the gas chamber is set to the initial temperature of the external temperature change curve; the gas chamber is meshed using structured meshing technology, the cylindrical gas chamber is divided into the upper and lower bottom surfaces and the side surfaces, a uniform equilateral triangle mesh is generated, and the contact edges are refined;

[0008] (2) Based on the model constructed in the previous step, the internal temperature field distribution characteristics of the SF6 gas chamber under different structural sizes in a variable temperature environment are obtained through finite element calculation, specifically including: adjusting the height, radius, wall thickness and internal SF6 gas pressure of the gas chamber in sequence; performing finite element calculation on the adjusted model to generate an internal average temperature data set corresponding to the external temperature changes under different structural sizes.

[0009] The present invention further provides a method for constructing a neural network model of a temperature data set generated based on finite element simulation, comprising the following steps:

[0010] (1) Preprocess the data set generated by finite element simulation, normalize the data and use it as a training set, define the input matrix and output matrix; set the neural network training parameters, including the maximum number of iterations and training accuracy

[0011] (2) Using the genetic particle swarm optimization algorithm to optimize the number of hidden layer neurons, learning rate, initial weights and bias of the neural network; setting the particle swarm size, inertia weight and learning factor of the genetic particle swarm optimization algorithm, and defining the fitness function as the predicted mean square error; by iteratively updating the particle speed and position, setting the genetic operation crossover probability and mutation probability, and generating the optimal parameter combination; assigning the optimal parameter combination to the neural network structure, and training to obtain a comprehensive performance parameter compensation model;

[0012] (3) The trained neural network model is verified on the test set to evaluate the prediction accuracy of the external temperature and the average temperature inside the SF6 gas chamber to ensure that the prediction accuracy meets the temperature compensation requirements in the actual application of the SF6 density relay. After the model is verified, it is applied to the actual temperature compensation process of the SF6 density relay. By inputting the current gas chamber structural parameters and external temperature in real time, the average temperature inside the gas chamber is accurately predicted to achieve precise temperature compensation.

[0013] The beneficial effects of the present invention are as follows: a data set is calculated in advance by physical field coupling software, and a neural network model is trained in advance using the data set; therefore, when actually compensating the SF6 density relay, it is only necessary to determine the SF6 gas chamber height, radius, wall thickness, initial gas pressure, external ambient temperature and initial gas chamber air temperature to accurately compensate for the internal average temperature of the SF6 gas chamber at each moment; compared with the existing human experience compensation method, the prediction result of the present invention is not affected by human experience, the compensation result is more accurate, and the compensation method is more convenient; compared with the existing function fitting method, the neural network model has better processing effect and higher accuracy for nonlinear fitting and parallel computing problems, thereby improving the analysis speed and efficiency of compensation. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A flow chart of a temperature compensation method for an SF6 density relay provided for the implementation of the present invention;

[0015] Figure 2 This is a schematic diagram of the SF6 gas chamber model structure;

[0016] Figure 3 This is the training flow chart of GA-PS0-RBF neural network. DETAILED DESCRIPTION

[0017] The practical basis of the present invention is that the simulation of the internal temperature field distribution characteristics of the SF6 gas chamber in a variable temperature environment based on a high-precision finite element simulation model includes the following steps:

[0018] (1) Geometric modeling of the SF6 gas chamber is performed in the multi-physics coupling simulation software; the material properties of the SF6 gas chamber are assigned, including thermal conductivity, specific heat capacity, density, etc. The material parameters are based on the physical properties of the SF6 gas and the gas chamber shell material; the external temperature is fitted using Fourier transform to obtain the external temperature change curve; in the solid and fluid heat transfer module, the outer wall temperature of the SF6 gas chamber is set to the external temperature change curve, and the initial temperature of the gas chamber is set to the initial temperature of the external temperature change curve; the gas chamber is meshed using structured meshing technology, the cylindrical gas chamber is divided into the upper and lower bottom surfaces and the side surfaces, a uniform equilateral triangle mesh is generated, and the contact edges are refined;

[0019] (2) Based on the model constructed in the previous step, the internal temperature field distribution characteristics of the SF6 gas chamber under different structural sizes in a variable temperature environment are obtained through finite element calculation, which specifically includes: adjusting the height, radius, wall thickness and internal SF6 gas pressure of the gas chamber in sequence; performing finite element calculation on the adjusted model to generate an average temperature data set inside the SF6 gas chamber corresponding to the external temperature changes under different structural sizes.

[0020] After the simulation model is built, the following steps are involved in building a neural network model based on the temperature data set generated by the finite element simulation:

[0021] (1) Preprocess the data set generated by finite element simulation, normalize the data and use it as a training set, and define the input matrix Where h is the height of the air chamber, r is the radius of the air chamber, and P is the initial pressure of the air chamber. is the wall thickness of the air chamber, T0 is the external temperature at the previous moment, T1 is the external temperature at this moment, T t0 is the average temperature inside the air chamber at the last moment. Output matrix Y={T t1}, T t1 The average temperature inside the air chamber at this moment; set the neural network training parameters, including the maximum number of iterations and training accuracy

[0022] (2) Using the genetic particle swarm optimization algorithm to optimize the number of hidden layer neurons, learning rate, initial weights and bias of the neural network; setting the particle dimension and number of the genetic particle swarm optimization algorithm, and defining the fitness function as the predicted mean square error; by iteratively updating the particle speed and position, setting the genetic operation crossover probability Px, mutation probability Pm, setting the neural network accuracy target, and generating the optimal parameter combination; assigning the optimal parameter combination to the neural network structure, and training to obtain a comprehensive performance parameter compensation model;

[0023] (3) The trained neural network model is verified on the test set to evaluate the prediction accuracy of the external temperature and the average temperature inside the SF6 gas chamber to ensure that the prediction accuracy meets the temperature compensation requirements in the actual application of the SF6 density relay. After the model is verified, it is applied to the actual temperature compensation process of the SF6 density relay. By inputting the current gas chamber structural parameters and external temperature in real time, the average temperature inside the gas chamber is accurately predicted to achieve precise temperature compensation.

[0024] The present invention will be further described below with reference to the accompanying drawings.

[0025] Figure 1 The figure shows a flow chart of the temperature compensation method for SF6 density relay. The specific process is as follows: 1) input SF6 gas chamber structural parameters; 2) establish SF6 gas chamber model; 3) finite element calculation; 4) use genetic particle swarm optimization algorithm to optimize the radial basis function neural network parameters; 5) construct GA-IPSO-RBF gas chamber compensation model; 6) output the average temperature of SF6 part.

[0026] Figure 2 The figure shows the results of the SF6 gas chamber model. The structural parameters include the chamber height, radius, wall thickness and the initial pressure inside the chamber.

[0027] Figure 3The figure shows the GA-IPSO-RBF training flow chart, and the specific process is as follows: 1) data normalization; 2) determine the RBF grid structure; 3) determine the PSO particle dimension and number; 4) define the fitness function; 5) initialize the particle velocity and position; 6) calculate the fitness function value; 7) update the particle velocity and position; 8) determine whether the number of iterations has been reached; 9) if the number of iterations has not been reached, perform crossover and mutation operations and re-execute processes 6 to 8; 10) if the number of iterations has been reached, update the RBF weights and configuration; 11) calculate the error; 12) update the weights and bias; 13) determine whether the accuracy requirements have been met; 14) if the accuracy requirements have not been met, re-execute processes 11 to 13; 15) if the accuracy requirements have been met, end.

Claims

1. A temperature compensation method for an SF6 density relay, characterized in that: The simulation of the internal temperature field distribution characteristics of the SF6 gas chamber in a variable temperature environment using a high-precision finite element simulation model includes the following steps: (1) Geometric modeling of the SF6 gas chamber is performed in the multi-physics coupling simulation software; the material properties of the SF6 gas chamber are assigned, including thermal conductivity, specific heat capacity, density, etc. The material parameters are based on the physical properties of the SF6 gas and the gas chamber shell material; the external temperature is fitted using Fourier transform to obtain the external temperature change curve; in the solid and fluid heat transfer module, the outer wall temperature of the SF6 gas chamber is set to the external temperature change curve, and the initial temperature of the gas chamber is set to the initial temperature of the external temperature change curve; the gas chamber is meshed using structured meshing technology, the cylindrical gas chamber is divided into the upper and lower bottom surfaces and the side surfaces, a uniform equilateral triangle mesh is generated, and the contact edges are refined; (2) Based on the model constructed in the previous step, the internal temperature field distribution characteristics of the SF6 gas chamber under different structural sizes in a variable temperature environment are obtained through finite element calculation, specifically including: adjusting the height, radius, wall thickness and internal SF6 gas pressure of the gas chamber in sequence; performing finite element calculation on the adjusted model to generate an internal average temperature data set corresponding to the external temperature changes under different structural sizes.

2. A temperature compensation method for an SF6 density relay according to claim 1, characterized in that: Building a neural network model based on the temperature dataset generated by finite element simulation involves the following steps: (1) Preprocess the data set generated by finite element simulation, normalize the data and use it as a training set, and define the input matrix Where h is the height of the air chamber, r is the radius of the air chamber, and P is the initial pressure of the air chamber. is the wall thickness of the air chamber, T0 is the external temperature at the previous moment, T1 is the external temperature at this moment, T t0 is the average temperature inside the air chamber at the last moment; output matrix Y = {T t1 }, T t1 is the average temperature inside the air chamber at this moment; set the neural network training parameters, including the maximum number of iterations and training accuracy; (2) Using the genetic particle swarm optimization algorithm to optimize the number of hidden layer neurons, learning rate, initial weights and bias of the neural network; setting the particle dimension and number of the genetic particle swarm optimization algorithm, and defining the fitness function as the predicted mean square error; by iteratively updating the particle speed and position, setting the genetic operation crossover probability Px, mutation probability Pm, setting the neural network accuracy target, and generating the optimal parameter combination; assigning the optimal parameter combination to the neural network structure, and training to obtain a comprehensive performance parameter compensation model; (3) The trained neural network model is verified on the test set to evaluate the prediction accuracy of the external temperature and the average temperature inside the SF6 gas chamber to ensure that the prediction accuracy meets the temperature compensation requirements in the actual application of the SF6 density relay. After the model is verified, it is applied to the actual temperature compensation process of the SF6 density relay. By inputting the current gas chamber structural parameters and external temperature in real time, the average temperature inside the gas chamber is accurately predicted to achieve precise temperature compensation.

Images