CN113435073B - GIS bus shell temperature measuring point arrangement optimization method - Google Patents

Abstract

A GIS bus shell temperature measuring point arrangement optimization method is suitable for being used by a power system. Constructing a three-dimensional space coordinate system and a GIS bus model according to the specific structure of the monitored GIS bus, and calculating a temperature field by utilizing finite element simulation; randomly selecting an initial value of an optimization algorithm, namely selecting coordinates of n measuring points in an established three-dimensional coordinate system and forming an initial vector i ₀ (ii) a Calculating the predicted value of the interpolation point by a Krigin space interpolation method and constructing a fitness function by utilizing the error between the predicted value and the simulated value; and solving the minimum value of the fitness function through a simulated annealing algorithm, wherein the position coordinate corresponding to the value is the arrangement position of the temperature measuring points. The method realizes the optimization of the GIS bus shell measuring point arrangement, can realize the optimization selection of the GIS temperature measuring points aiming at the problems of more GIS temperature measuring points and low precision in reality, and provides reliable guarantee for the state monitoring and evaluation of the actual engineering.

Description

A method for optimizing the layout of GIS busbar shell temperature measuring points

Technical field:

The invention relates to a method for optimizing the arrangement of temperature measuring points, in particular to a method for optimizing the arrangement of temperature measuring points for a GIS busbar casing used in a power system.

Background technique:

With the continuous growth of my country's electricity demand and the development and popularization of high-voltage electrical appliances, the application of gas-insulated metal-enclosed switch GIS in high-voltage substations and UHV transmission projects has increased dramatically. GIS equipment has strict processing technology, integrated molding, small footprint, and SF ₆ gas makes it have good breaking capacity. However, with the increasing service life, defects such as insulation aging or poor contact are gradually exposed. These defects often first appear as abnormal changes in gas temperature and shell temperature. Therefore, it is of great significance to carry out comprehensive temperature online monitoring on GIS, give early warning to abnormalities and defects and ensure its reliable operation.

Due to the unique structure of GIS, the direct temperature measurement of the conductor needs to open holes on the shell or directly install sensors inside, which will have adverse effects on the insulation, temperature field, electric field, etc. of the GIS, so it is difficult to directly measure the conductor temperature effectively. At present, the temperature measurement methods for GIS are direct monitoring of the shell, including infrared temperature measurement technology, fiber Bragg grating temperature measurement technology, and contact thermocouple and thermistor temperature measurement technology. Infrared temperature measurement is not in direct contact with the conductor and the shell, but the accuracy is low and is affected by factors such as the emissivity of the metal surface and the density of SF ₆ gas. The cost of grating optical fiber sensor is relatively high, and due to the particularity of GIS structure, it has not been widely used in GIS equipment temperature monitoring. In this paper, the contact temperature measurement device is used to monitor by adhering the sensor on the surface of the shell.

At present, the selection of temperature measurement points is mostly limited to the shell corresponding to the conductor contact. There is no relevant information about the specific location and basis for the selection of temperature measurement points, which is not helpful for the selection of on-site sensor installation and layout locations. It is very necessary to choose the location of the temperature measurement point of the shell, which can greatly improve the reliability of GIS temperature monitoring and provide assistance for on-site operation and maintenance personnel.

Contents of the invention

The purpose of the present invention is to provide a simple step, fast calculation speed, solve the problem of selecting the GIS shell temperature measurement point from the perspective of three-dimensional space, consider the overall layout structure of the GIS busbar, combine Kriging space interpolation method and simulated annealing optimization algorithm An optimization method for the layout of GIS busbar shell temperature measuring points for selecting the best temperature measuring points.

In order to solve the above-mentioned technical problems, a kind of GIS busbar housing temperature measuring point arrangement optimization method of the present invention, its specific steps are as follows:

Step S1: Construct the 3D coordinate system and the 3D model of the GIS busbar, analyze the magnetic field-flow field-temperature field of the multi-physics field coupled 3D model, and use the finite element calculation to obtain the overall temperature distribution of the GIS busbar shell in the normal state. Busbars of different voltage levels are divided into sections according to standard units and the sections are selected for layout of measuring points. The three-dimensional coordinate definition domain of this section is the feasible solution space;

Step S2: Select m points whose interpolation temperature T' is unknown in the measuring point layout section of the 3D model, and use the temperature information of the temperature measuring points of n sensors to be arranged in the 3D measuring point layout section of the GIS bus to assign influence weights, Introduce kriging space interpolation method to interpolate the interpolation temperature T' of m points in turn, and then calculate the square root mean error between the interpolation temperature T' of m points and the temperature T calculated by finite element simulation;

Step S3: Randomly select n points in the feasible solution space as the temperature measuring points of the sensors to be arranged, and form an initial vector i ₀ with the coordinates of all points, and input the initial vector i ₀ into the simulated annealing algorithm of MATLAB as the initial value, Taking the square root mean error as the fitness function f(i) of the simulated annealing algorithm, given the remaining initial parameters of the simulated annealing algorithm including the initial simulated annealing temperature T ₀ and the length L ₀ of the Markov chain, the fitness function f(i ), the minimum value minf(i) of the fitness function f(i) is the optimal solution vector, and the n coordinates included are the layout of the optimal temperature measuring points on the GIS busbar enclosure.

Use 3D modeling software such as SolidWorks to construct the busbar structure of the sensor GIS to be arranged in 3D space coordinates, obtain the coordinates of each point of the GIS shell, and import the model into the multi-physics finite element simulation software COMSOL, and perform a magnetic field-fluid field analysis on it. — Multi-field coupled finite element analysis of the temperature field, with the actual operation conditions on site as the boundary conditions, the overall temperature distribution of the GIS enclosure under normal operating conditions is calculated. Since the structures of the GIS busbars of different voltage levels are slightly different, the GIS busbars are calculated according to The standard unit is divided into sections, and then the section where the sensor needs to be arranged to measure the temperature of the shell is determined, and the feasible solution space is determined by the three-dimensional space coordinate definition domain of the section, and then n sensors to be arranged in the feasible solution space are arbitrarily randomly selected The temperature measurement points (xi _, y _i , z _i ), i=1,2,…n. The initial vector i ₀ is constructed as the initial value of the simulated annealing optimization algorithm, and the distribution characteristics of the temperature field of the busbar at different voltage levels are different, The number n of initial measuring points should be selected according to the corresponding GIS voltage level.

Randomly select m temperature points to be interpolated in the three-dimensional measuring point layout section, and according to the principle of spatial interpolation method, the observed values Z _i of the temperature attributes of several discrete points (xi _{, y} , _zi ) in the known space Estimate the temperature value of any point (x _{, y} , z) in space under the condition of = (xi, y i, zi ₎ ; due to the complex structure of GIS, the distance distribution between each measuring point is uneven and in three dimensions In the structure, the kriging interpolation method based on the distribution of influence weights is introduced, and the kriging interpolation formula is as follows:

是点(x,y,z)处的估计值，即Z₀＝Z(x₀,y₀,z₀)；λ_i是权重系数；克里金插值法利用数据加权求和来估算未知点的值，权重系数是能够满足点(x₀,y₀,z₀)处的估算值

与真实值Z₀的差最小的一套最优系数；in

is the estimated value at the point (x,y,z), that is, Z ₀ =Z(x ₀ ,y ₀ ,z ₀ ); λ _i is the weight coefficient; the Kriging interpolation method uses the weighted sum of the data to estimate the unknown point The value of the weight coefficient is the estimated value at the point (x ₀ , y ₀ , z ₀ ) that can satisfy

A set of optimal coefficients with the smallest difference from the true value Z ₀ ;

According to the principle of kriging interpolation, weight coefficients are assigned to the temperature information of n sensor arrangement points, and the temperature values to be interpolated T'(x' _j ,y' _j ,z' _j ) of m points are predicted sequentially, j=1 ,2,...m.; compare the interpolated temperature information of m measurement points with the simulated temperature T(x _j ,y _j ,z _j ), and calculate the mean square error between the two, so as to construct the simulated annealing algorithm fitness function;

The fitness function sets constraints:

(1) According to the positions of different sections in the three-dimensional space coordinate system, set the feasible range of sensor arrangement point coordinates;

若d＜d₀，d₀为距离的阈值，则给该解对应的适应度函数值加一个惩罚值以滤除该解；(2) In order to avoid the distance between the n layout points being too close and the interpolation results affected by the mutual interference between the sensors, it is necessary to set constraints on the uniformity of the layout points; for a solution i=(x ₁ ,y ₁ ,z ₁ , x ₂ ,y ₂ ,z ₂ ,···,x _n ,y _n ,z _n ), let the center of gravity coordinates be (x _w ,y _w ,z _w ), w∈(1,n)w belongs to 1- parameter in n, defining a scatter distance

If d<d ₀ , d ₀ is the threshold value of the distance, then add a penalty value to the fitness function value corresponding to the solution to filter out the solution;

Based on the above analysis, the fitness function is expressed as:

为m个待插点的插值温度，T(x_j,y_j,z_j)为m个待插点的仿真温度，Z(x_i,y_i,z_i)为n个优化测点的温度。in,

is the interpolation temperature of m points to be inserted, T(x _j ,y _j ,z _j ) is the simulation temperature of m points to be inserted, Z( _xi ,y _i , _zi ) is the temperature of n optimized measuring points .

The fitness function value f(i ₀ ) of the initial vector i ₀ formed by n initially randomly selected temperature measuring points, at this time, the optimal solution of the simulated annealing algorithm is i=i ₀ , and then a new solution generation rule is generated Random perturbation to obtain a new solution j of the same dimension as the initial vector i ₀ in the feasible solution space, which is a vector composed of n different temperature measuring point coordinates of sensors to be arranged, and then judge whether to accept the new solution j according to the Metropolis criterion , repeatedly perform L ₀ times of new solution generation and judgment, and obtain the optimal solution under the chain length of L ₀ , judge whether the optimal solution of continuous C steps is satisfied, and output the optimal solution if it is satisfied, and each point in the optimal solution vector The coordinates are the layout positions of the best temperature measuring points in this section; otherwise, continue to iteratively solve.

Specifically, the initial vector i ₀ =[(x ₀₁ ,y ₀₁ ,z ₀₁ ),(x 02 ,y 02 ,z 02 ),(x ₀₂ ,y ₀₂ ,z ₀₂ ),…,( x _0n ,y _0n ,z _0n )] into the fitness function f(i) composed of the square root mean error of the interpolated temperature and the simulated temperature, to obtain the initial value f(i ₀ ), and the initial optimal solution is i ₀ ;

The new solution j is generated according to the following rules:

对于每一个i∈[1,n]，计算

及

然后将更新后的坐标构成新解向量j；First generate a set of random numbers (a ₁ ,a ₂ ,…a _n ), where a _i obeys the normal distribution N(0,1); then calculate (b ₁ ,b ₂ ,…b _n ), where

For each i∈[1,n], compute

and

Then the updated coordinates constitute a new solution vector j;

Whether to accept the new solution is judged by the Metropolis criterion:

时，接受新解j，此时最优解i＝j，反之则拒绝j，此时最优解仍为i；If f(j)-f(i)<0, let i=j, otherwise accept the new solution j according to the probability, namely

, accept the new solution j, at this time the optimal solution i=j, otherwise reject j, at this time the optimal solution is still i;

If the new solution j is accepted, according to the Kriging interpolation formula, it is necessary to determine a new weight coefficient distribution to obtain the interpolation temperature of the m points to be interpolated under the new solution, so a second optimization is performed on the weight coefficient λ;

In the case that the new solution j has been determined, λ ₀ = (λ ₀₁ ,λ ₀₂ ,…,λ _0n ) is substituted into the fitness function f(λ) as the initial vector, and the interpolation temperature that can make the m points to be interpolated The distribution of weight coefficients with the smallest mean square error with the simulated temperature, namely:

Repeat the process of generating and judging new solutions L _k times to obtain an optimal solution under the chain length L _k ; judge whether the algorithm satisfies the termination criterion, the termination criterion should take into account both optimization performance and efficiency, in order to avoid too many unnecessary searches Therefore, the threshold discriminant method is adopted, that is, if the optimal solution remains unchanged during the cooling period of continuous C steps, it is approximately considered to be convergent; when the algorithm meets the termination condition, the optimal solution is output and the algorithm is stopped. When it is not satisfied, the number of iterations k=k+1, the decay function of the control parameter T is updated to T _k+1 , and the Markov chain is L _k+1 ; in order to improve the efficiency of the algorithm while ensuring the accuracy of the solution and the stability of the simulated annealing algorithm , the attenuation function of the control parameter T, that is, the cooling function T _k is T _k+1 = α T _k , k = 0,1,2,3..., α is usually 0.95, then the temperature of the kth iteration is T _k = T ₀ ×0.95 ^k ;

When the algorithm is terminated, the coordinates of each point in the optimal solution vector are the optimization results of sensor measuring point layout

Beneficial effect:

The temperature measurement point layout optimization method mentioned in the present invention aims at the problem that the sensors installed on the GIS field can not determine the temperature influence range and the existing GIS temperature sensor layout scheme is to optimize the number of sensors, and a GIS bus is proposed. Optimization method for the layout of the shell temperature measuring points;

The present invention is applicable to compact and standard GIS, and uses simulated annealing algorithm to optimize the position of the measuring point for the measuring point layout section composed of the standard unit of the GIS bus, with high calculation accuracy and fast convergence speed, and can quickly obtain the best temperature measuring point. The location is arranged to save time and cost, and the implementation is simple. Combined with the temperature sensor, the real-time monitoring of the shell temperature is realized, which is convenient for the on-site operation and maintenance personnel to track the health status of the equipment, and provides a basis for state identification based on temperature difference characteristics.

Description of drawings

Fig. 1 is the flow chart of the method for optimizing the arrangement of GIS shell temperature measurement points of the present invention;

Figure 2 is a structural diagram of the 500kVGIS busbar;

Figure 3 is a structural diagram of the 220kVGIS busbar;

Fig. 4 is the flowchart of interpolation method and simulated annealing algorithm;

Figure 5 is the layout of temperature measuring points for the shell of the 500kV GIS bus section.

Detailed ways

Below in conjunction with accompanying drawing, technical scheme of the present invention is described in further detail:

As shown in Figure 1, a kind of GIS busbar housing temperature measuring point arrangement optimization method of the present invention, its steps are as follows:

Step S1: Construct the 3D coordinate system and the 3D model of the GIS busbar, analyze the magnetic field-flow field-temperature field of the multi-physics field coupled 3D model, and use the finite element calculation to obtain the overall temperature distribution of the GIS busbar shell in the normal state. The busbars of different voltage levels are divided into sections according to standard units and the measurement point layout section is selected. The three-dimensional coordinate definition domain of this section is the feasible solution space, and then n points are randomly selected in the feasible solution space as the temperature measurement points of the sensors to be arranged. Points, the coordinates of all points constitute an initial vector i ₀ , which is used as the initial value of the simulated annealing optimization algorithm; use 3D modeling software such as SolidWorks to construct the busbar structure of the sensor GIS to be arranged in the 3D space coordinates, and obtain the GIS shell for each The coordinates of one point are imported into the multi-physics finite element simulation software COMSOL, and the magnetic field-fluid field-temperature field multi-field coupling finite element analysis is performed on it. The actual operation conditions on site are used as boundary conditions, and the calculation results are obtained under normal operating conditions. Under the overall temperature distribution of the GIS shell, because the structure of the GIS bus bar with different voltage levels is slightly different, the GIS bus bar is divided into sections according to its standard unit, and then the section where the sensor needs to be arranged to measure the shell temperature is determined, and the section is used The coordinate definition domain of the three-dimensional space determines the feasible solution space, and then arbitrarily randomly selects the temperature measurement points (xi _, y _i , z _i ) of n sensors to be arranged in the feasible solution space, i=1,2,…n. The initial vector i ₀ is used as the initial value of the simulated annealing optimization algorithm, and the bus temperature field distribution characteristics of different voltage levels are different, and the number of initial measuring points n should be selected according to the corresponding GIS voltage level;

Step S2: Introduce the Kriging spatial interpolation method, select m points whose interpolation temperature T' is unknown in the 3D measuring point layout section, and use the temperature measuring points of n sensors to be arranged in the 3D measuring point layout section of the GIS bus The temperature information assigns influence weights, interpolates the interpolated temperature T' of m points in turn, and then takes the square root mean error between the interpolated temperature T' of m points and the temperature T calculated by finite element simulation as the fitness of the simulated annealing algorithm function f(i);

Step S3: Use the simulated annealing algorithm to find the optimal solution, randomly select n points in the feasible solution space as the temperature measurement points of the sensor to be arranged, and form an initial vector i ₀ with the coordinates of all points, and use it as the simulated annealing algorithm Initial value; take the square root mean error as the fitness function f(i) of the simulated annealing algorithm, given the other initial parameters of the simulated annealing algorithm including the initial simulated annealing temperature T ₀ and the length L ₀ of the Markov chain, calculate the fitness function The minimum value minf(i) of f(i), the fitness function value f(i ₀ ) of the initial vector i ₀ formed by n initially randomly selected temperature measuring points, the optimal solution of the simulated annealing algorithm at this time is i = i ₀ , and then generate a random disturbance according to the new solution generation rule, and obtain a new solution j in the same dimension as the initial vector i ₀ in the feasible solution space, which is composed of the temperature measuring point coordinates of n different sensors to be arranged Then judge whether to accept the new solution j according to the Metropolis criterion, repeat the generation and judgment of the new solution L ₀ times, and obtain the optimal solution with a chain length of L ₀ , and judge whether the optimal solution of continuous C steps is satisfied, and then Output the optimal solution, and the coordinates of each point in the optimal solution vector are the layout positions of the best temperature measuring points in this section; otherwise, continue to iteratively solve.

Combined with attached drawings 2 and 3, the structure diagrams of 220kV and 500kV GIS busbars, the busbars of high-voltage gas insulated switchgear are divided into three-phase common type and three-phase split type according to the different voltage levels, and most of the GIS busbars below 220kV are three-phase Phase communal type, the ABC three-phase is arranged in an inverted triangle, the AC two-phase is on the top and the B-phase is on the bottom. For the GIS busbar of 500kV and above, in order to ensure its insulation performance, most of them are three-phase split type; different structures have relatively large temperature field differences. Obviously, the number of sensors is different. In order to improve the calculation efficiency, the buses of different voltage levels in GIS are modeled separately and the temperature measurement point layout optimization calculation is carried out.

Combined with the flow chart of the interpolation method and simulated annealing algorithm in Figure 4, randomly select m temperature points to be interpolated in the three-dimensional measuring point layout section, and according to the principle of the spatial interpolation method, several discrete points in the known space (x _i , y _i , z _i ) Under the condition that the observed value of the temperature attribute Z _i = (xi _, y _i , _zi ), estimate the temperature value at any point (x, y, z) in space; due to the complex structure of GIS, The distance distribution between each measurement point is uneven and in a three-dimensional structure, so the kriging interpolation method based on the distribution of influence weights is introduced. The kriging interpolation formula is as follows:

是点(x,y,z)处的估计值，即Z₀＝Z(x₀,y₀,z₀)；λ_i是权重系数；克里金插值法利用数据加权求和来估算未知点的值，权重系数是能够满足点(x₀,y₀,z₀)处的估算值

与真实值Z₀的差最小的一套最优系数；in

is the estimated value at the point (x,y,z), that is, Z ₀ =Z(x ₀ ,y ₀ ,z ₀ ); λ _i is the weight coefficient; the Kriging interpolation method uses the weighted sum of the data to estimate the unknown point The value of the weight coefficient is the estimated value at the point (x ₀ , y ₀ , z ₀ ) that can satisfy

A set of optimal coefficients with the smallest difference from the true value Z ₀ ;

According to the principle of kriging interpolation, weight coefficients are assigned to the temperature information of n sensor arrangement points, and the temperature values to be interpolated T'(x' _j ,y' _j ,z' _j ) of m points are predicted sequentially, j=1 ,2,...m.; compare the interpolated temperature information of m measurement points with the simulated temperature T(x _j ,y _j ,z _j ), and calculate the mean square error between the two, so as to construct the simulated annealing algorithm fitness function;

The fitness function sets constraints:

(1) According to the positions of different sections in the three-dimensional space coordinate system, set the feasible range of sensor arrangement point coordinates;

若d＜d₀，d₀为距离的阈值，则给该解对应的适应度函数值加一个惩罚值以滤除该解；(2) In order to avoid the distance between the n layout points being too close and the interpolation results affected by the mutual interference between the sensors, it is necessary to set constraints on the uniformity of the layout points; for a solution i=(x ₁ ,y ₁ ,z ₁ , x ₂ ,y ₂ ,z ₂ ,···,x _n ,y _n ,z _n ), let the barycentric coordinates be (x _w ,y _w ,z _w ), w∈(1,n), define a dispersion distance

If d<d ₀ , d ₀ is the threshold value of the distance, then add a penalty value to the fitness function value corresponding to the solution to filter out the solution;

Based on the above analysis, the fitness function can be expressed as:

为m个待插点的插值温度，T(x_j,y_j,z_j)为m个待插点的仿真温度，Z(x_i,y_i,z_i)为n个优化测点的温度；in,

is the interpolation temperature of m points to be inserted, T(x _j ,y _j ,z _j ) is the simulation temperature of m points to be inserted, Z( _xi ,y _i , _zi ) is the temperature of n optimized measuring points ;

Given the initial temperature T ₀ of the simulated annealing algorithm and the initial Markov chain length L ₀ , then the initial vector i ₀ =[(x ₀₁ ,y ₀₁ ,z ₀₁ ),(x ₀₂ ,y ₀₂ ,z ₀₂ ),…,(x _0n ,y _0n ,z _0n )] into the fitness function f(i) composed of the square root mean error of the interpolated temperature and the simulated temperature, and obtain Initial value f(i ₀ ), and the initial optimal solution is i ₀ ;

The new solution is generated according to the following rules:

对于每一个i∈[1,n]，计算

及

然后将更新后的坐标构成新解向量j；First generate a set of random numbers (a ₁ ,a ₂ ,…a _n ), where a _i obeys the normal distribution N(0,1); then calculate (b ₁ ,b ₂ ,…b _n ), where

For each i∈[1,n], compute

and

Then the updated coordinates constitute a new solution vector j;

Whether to accept the new solution is judged by the Metropolis criterion:

时，接受新解j，此时最优解i＝j，反之则拒绝j，此时最优解仍为i；If f(j)-f(i)<0, let i=j, otherwise accept the new solution j according to the probability, namely

, accept the new solution j, at this time the optimal solution i=j, otherwise reject j, at this time the optimal solution is still i;

If the new solution j is accepted, according to the kriging interpolation formula, it is necessary to determine a new weight coefficient distribution to obtain the interpolation temperature of the m points to be interpolated under the new solution, so a second optimization is performed on the weight coefficient λ.

When the new solution has been determined, λ ₀ =(λ ₀₁ ,λ ₀₂ ,…,λ _0n ) is substituted into the fitness function f(λ) as the initial vector, and the interpolation temperature and The distribution of weight coefficients with the smallest mean square error between simulation temperatures, namely:

Repeat the process of generating and judging new solutions L _k times to obtain an optimal solution under the chain length L _k ; judge whether the algorithm satisfies the termination criterion, the termination criterion should take into account both optimization performance and efficiency, in order to avoid too many unnecessary searches Therefore, the threshold discriminant method is adopted, that is, if the optimal solution remains unchanged during the cooling period of continuous C steps, it is approximately considered to be convergent; when the algorithm meets the termination condition, the optimal solution is output and the algorithm is stopped. When it is not satisfied, the number of iterations k=k+1, the decay function of the control parameter T is updated to T _k+1 , and the Markov chain is L _k+1 ; in order to improve the efficiency of the algorithm while ensuring the accuracy of the solution and the stability of the simulated annealing algorithm , the attenuation function of the control parameter T, that is, the cooling function T _k is T _k+1 = α T _k , k = 0,1,2,3..., α is usually 0.95, then the temperature of the kth iteration is T _k = T ₀ ×0.95 ^k ;

When the algorithm is terminated, the coordinates of each point in the optimal solution vector are the optimization results of sensor measuring point layout.

Combined with the layout of the 500kV GIS bus shell temperature measuring points shown in Figure 5, establish a three-dimensional space coordinate system for the bus section, simplify the elements in the model that have little influence on the temperature distribution, and construct a multi-physics model of the electromagnetic field, thermal field, and fluid field The field coupling finite element model is used to obtain the temperature distribution of the shell during the normal operation of the GIL, and the arrangement of the measuring points is optimized through the interpolation method and the simulated annealing algorithm to obtain the layout of the space temperature measuring points;

In this embodiment, one air chamber of the busbar is 18m long. According to the characteristics of its temperature distribution, that is, the axial temperature difference in the air chamber is small, and the radial section presents a trend of higher up and lower down, an initial sensor for the air chamber section of the busbar is set The number n of layouts is 24, and the number m of points to be interpolated is 30;

The parameters of the simulated annealing algorithm are set as follows: the initial temperature T ₀ is 1000°C, then the cooling function is T _k =1000×0.95 ^k ; the Markov chain length L ₀ =L _k =30 remains unchanged; the optimal solution in the termination criterion is not Change the number of steps to continuous C=L ₀ ＝30 steps; solve the fitness function

The coordinate vector set at the minimum value is the location of the final shell temperature measuring point.

According to the results of the above theoretical analysis and example calculation, it is effective and feasible to optimize the arrangement of the shell temperature measuring points by this method.

In this embodiment, a 500kV GIS bus space coordinate model is established, the shell temperature is calculated by finite element, the initial measuring point position is selected to estimate the unknown point temperature according to the interpolation method, and then the optimal measuring point position coordinates are obtained by substituting the simulated annealing algorithm. In this way, the optimization of the arrangement of temperature measuring points for the shell of the GIS bus section can be realized.

In this embodiment, through the division and selection of the GIS structure, it is beneficial to reduce the complexity of the algorithm and shorten the calculation time. Through the spatial interpolation method, the basis for the arrangement of the temperature measurement points in the spatial position can be provided.

Claims (4)

1. a GIS busbar housing temperature measurement point layout optimization method is characterized in that concrete steps are as follows: Step S1: Construct the 3D coordinate system and the 3D model of the GIS busbar, analyze the magnetic field-flow field-temperature field of the multi-physics field coupled 3D model, and use the finite element calculation to obtain the overall temperature distribution of the GIS busbar shell in the normal state. Busbars of different voltage levels are divided into sections according to standard units and the sections are selected for layout of measuring points. The three-dimensional coordinate definition domain of this section is the feasible solution space; Step S2: Select m points whose interpolation temperature T' is unknown in the measuring point layout section of the 3D model, and use the temperature information of the temperature measuring points of n sensors to be arranged in the 3D measuring point layout section of the GIS bus to assign influence weights, Introduce kriging space interpolation method to interpolate the interpolation temperature T' of m points in turn, and then calculate the square root mean error between the interpolation temperature T' of m points and the temperature T calculated by finite element simulation; Step S3: Randomly select n points in the feasible solution space as the temperature measuring points of the sensors to be arranged, and form an initial vector i ₀ with the coordinates of all points, and input the initial vector i ₀ into the simulated annealing algorithm of MATLAB as the initial value, Taking the square root mean error as the fitness function f(i) of the simulated annealing algorithm, given the remaining initial parameters of the simulated annealing algorithm including the initial simulated annealing temperature T ₀ and the length L ₀ of the Markov chain, the fitness function f(i ), the minimum value minf(i) of the fitness function f(i) vector is the optimal solution vector, and the n coordinates included are the layout of the optimal temperature measuring points on the GIS busbar shell; Randomly select m temperature points to be interpolated in the three-dimensional measuring point layout section, and according to the principle of spatial interpolation method, the observed values Z _i of the temperature attributes of several discrete points (xi _{, y} , _zi ) in the known space Estimate the temperature value of any point (x _{, y} , z) in space under the condition of = (xi, y i, zi ₎ ; due to the complex structure of GIS, the distance distribution between each measuring point is uneven and in three dimensions In the structure, the kriging interpolation method based on the distribution of influence weights is introduced, and the kriging interpolation formula is as follows:

in

is the estimated value at the point (x,y,z), that is, Z ₀ =Z(x ₀ ,y ₀ ,z ₀ ); λ _i is the weight coefficient; the Kriging interpolation method uses the weighted sum of the data to estimate the unknown point The value of the weight coefficient is the estimated value at the point (x ₀ , y ₀ , z ₀ ) that can satisfy

A set of optimal coefficients with the smallest difference from the true value Z ₀ ; According to the principle of kriging interpolation, weight coefficients are assigned to the temperature information of n sensor arrangement points, and the temperature values to be interpolated T'(x' _j ,y' _j ,z' _j ) of m points are predicted sequentially, j=1 ,2,...m.; compare the interpolated temperature information of m measurement points with the simulated temperature T(x _j ,y _j ,z _j ), and calculate the mean square error between the two, so as to construct the simulated annealing algorithm fitness function; The fitness function sets constraints: (1) According to the positions of different sections in the three-dimensional space coordinate system, set the feasible range of sensor arrangement point coordinates; (2) In order to avoid the distance between the n layout points being too close and the interpolation results affected by the mutual interference between the sensors, it is necessary to set constraints on the uniformity of the layout points; for a solution i=(x ₁ ,y ₁ ,z ₁ , x ₂ ,y ₂ ,z ₂ ,···,x _n ,y _n ,z _n ), let the center of gravity coordinates be (x _w ,y _w ,z _w ), w∈(1,n)w belongs to 1- parameter in n, defining a scatter distance

If d<d ₀ , d ₀ is the threshold value of the distance, then add a penalty value to the fitness function value corresponding to the solution to filter out the solution; Based on the above analysis, the fitness function is expressed as:

in,

is the interpolation temperature of m points to be inserted, T(x _j ,y _j ,z _j ) is the simulation temperature of m points to be inserted, Z( _xi ,y _i , _zi ) is the temperature of n optimized measuring points . 2. a kind of GIS busbar housing temperature measuring point layout optimization method according to claim 1 is characterized in that utilize SolidWorks three-dimensional modeling software to build the busbar structure of sensor GIS to be arranged in three-dimensional space coordinates, obtain the GIS housing every point Coordinates, and import the model into the multi-physics finite element simulation software COMSOL, conduct multi-field coupling finite element analysis of magnetic field-fluid field-temperature field, and use the actual operation conditions on site as boundary conditions to calculate the GIS under normal operating conditions The overall temperature distribution of the shell, because the GIS busbar structure of different voltage levels is slightly different, the GIS busbar is divided into sections according to its standard unit, and then the section where the sensor needs to be arranged to measure the shell temperature is determined, and the three-dimensional The space coordinate definition domain determines the feasible solution space, and then arbitrarily randomly selects the temperature measurement points (xi _, y _i , z _i ) of n sensors to be arranged in the feasible solution space, i=1,2,…n. Construct the initial vector i ₀ is used as the initial value of the simulated annealing optimization algorithm, and the bus temperature field distribution characteristics of different voltage levels are different, and the initial number of measuring points n should be selected according to the corresponding GIS voltage level. 3. a kind of GIS busbar casing temperature measuring point layout optimization method according to claim 1 is characterized in that: the fitness function value f(i ₀ of the initial vector i ₀ formed by the temperature measuring points selected at n initial ), then the optimal solution of the simulated annealing algorithm is i=i ₀ , and then a random disturbance is generated according to the new solution generation rule, and a new solution j of the same dimension as the initial vector i ₀ in the feasible solution space is obtained, that is, n The vector composed of the temperature measuring point coordinates of different sensors to be arranged, and then judge whether to accept the new solution j according to the Metropolis criterion, repeat the generation and judgment of the new solution for L ₀ times, and obtain the optimal solution with a chain length of L ₀ , judge Whether the optimal solution of continuous steps C is satisfied, the optimal solution is output if it is satisfied, and the coordinates of each point in the optimal solution vector are the layout positions of the best temperature measuring points in this section; otherwise, continue to iteratively solve. 4. A method for optimizing the arrangement of GIS busbar housing temperature measurement points according to claim 3, characterized in that: specifically, the initial vector i ₀ =[ (x ₀₁ ,y ₀₁ ,z ₀₁ ),(x ₀₂ ,y ₀₂ ,z ₀₂ ),…,(x _0n ,y _0n ,z _0n )] is substituted into the fitness degree consisting of the square root mean error between the interpolated temperature and the simulated temperature In the function f(i), the initial value f(i ₀ ) is obtained, and the initial optimal solution is i ₀ ; The new solution j is generated according to the following rules: First generate a set of random numbers (a ₁ ,a ₂ ,…a _n ), where a _i obeys the normal distribution N(0,1); then calculate (b ₁ ,b ₂ ,…b _n ), where

For each i∈[1,n], compute

and

Then the updated coordinates constitute a new solution vector j; Whether to accept the new solution is judged by the Metropolis criterion: If f(j)-f(i)<0, let i=j, otherwise accept the new solution j according to the probability, namely

, accept the new solution j, at this time the optimal solution i=j, otherwise reject j, at this time the optimal solution is still i; If the new solution j is accepted, according to the Kriging interpolation formula, it is necessary to determine a new weight coefficient distribution to obtain the interpolation temperature of the m points to be interpolated under the new solution, so a second optimization is performed on the weight coefficient λ; In the case that the new solution j has been determined, λ ₀ = (λ ₀₁ ,λ ₀₂ ,…,λ _0n ) is substituted into the fitness function f(λ) as the initial vector, and the interpolation temperature that can make the m points to be interpolated The distribution of weight coefficients with the smallest mean square error with the simulated temperature, namely:

Repeat the process of generating and judging new solutions L _k times to obtain an optimal solution under the chain length L _k ; judge whether the algorithm satisfies the termination criterion, the termination criterion should take into account both optimization performance and efficiency, in order to avoid too many unnecessary searches Therefore, the threshold discriminant method is adopted, that is, if the optimal solution remains unchanged during the cooling period of continuous C steps, it is approximately considered to be convergent; when the algorithm meets the termination condition, the optimal solution is output and the algorithm is stopped. When it is not satisfied, the number of iterations k=k+1, the decay function of the control parameter T is updated to T _k+1 , and the Markov chain is L _k+1 ; in order to improve the efficiency of the algorithm while ensuring the accuracy of the solution and the stability of the simulated annealing algorithm , the attenuation function of the control parameter T, that is, the cooling function T _k is T _k+1 = α T _k , k = 0,1,2,3..., α is usually 0.95, then the temperature of the kth iteration is T _k = T ₀ ×0.95 ^k ; When the algorithm is terminated, the coordinates of each point in the optimal solution vector are the optimization results of sensor measuring point layout.

Images