Method for calculating motor performance stray loss

Abstract

The application discloses a kind of calculation methods of motor efficiency stray loss, comprising the following steps: a) establishing dependent variable is quantile regression model of stray loss: b) set quantile value, realize the estimation of the parameter vector corresponding to different quantiles by minimizing the residual of stray loss real value and estimated value, obtain the fitting curve cluster of stray loss sample data;C) the goodness-of-fit test is carried out to the curve in step b), and the combination of the best curve or curve segment of fitting effect is screened out, to determine the motor efficiency stray loss thereof.The application starts from quantile regression method, obtains different curves by selecting different quantiles, selects the curve or curve segment most consistent with fitting standard by the test of goodness-of-fit, can reduce the influence of abnormal point on curve.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of motor efficiency calculation, and particularly relates to a motor efficiency stray loss calculation method. BACKGROUND

[0002] At present, motors are widely used in various fields such as industry, commerce, and important national departments. On the one hand, motors are used as generators to provide power for various devices, and on the other hand, motors are used as electric motors to output work externally. However, the efficiency of the motor needs to be improved. The average efficiency of small and medium-sized motors in China is 87%, and the average design efficiency of fans and pumps is 75%. The operating efficiency of motor systems and motors and driving equipment is about 20% lower than that of foreign advanced technology. The long-term energy-saving special plan of the state clearly points out that "motor system energy saving" is one of the ten key projects. The motor efficiency standard evaluation methods of various countries are roughly the same. The current method adopted in China is "GB / T1032-2012 Three-phase Asynchronous Motor Test Method". The analysis basis mainly relies on test data under five different load conditions. With the optimization and improvement of motor experimental methods, rich experimental data can be easily obtained. The present application aims to improve the motor efficiency evaluation method in the standard "GB / T1032-2012 Three-phase Asynchronous Motor Test Method" based on these more abundant test data. SUMMARY

[0003] The application aims to provide a motor efficiency stray loss calculation method which is beneficial to reduce test errors and improve the calculation accuracy of stray loss.

[0004] The application aims to achieve the above-mentioned application by the following technical scheme: a motor efficiency stray loss calculation method, comprising the following steps:

[0005] a) establishing a quantile regression model with stray loss as the dependent variable:

[0006]

[0007] T is the estimated value of stray loss, T 2 represents the shaft torque, τ represents the quantile, β is the general term of β1(τ) and β0(τ), and represents the parameter vector;

[0008] b) setting the quantile value, estimating the parameter vector corresponding to different quantiles by minimizing the residual of the true value and the estimated value of stray loss, and obtaining the fitting curve cluster of stray loss sample data;

[0009] c) performing goodness-of-fit test on the curve in step b), screening out the curve with the best fitting effect, and determining the motor efficiency stray loss according to the curve;

[0010] Or, the curve in step b) is segmented, and the goodness of fit test is performed on each segment to obtain the curve segment with the best fitting effect in each interval, and the motor performance stray loss is determined by the curve segment.

[0011] In step b), the fitting curve of the stray loss sample data is obtained by the following method

[0012] b1) determine the value of quantile τ;

[0013] b2) find the quantile point corresponding to the quantile τ: (T m 2 , P Sm );

[0014] b3) for a given N observation samples, the parameter vector β is evaluated by the following formula:

[0015]

[0016] wherein, is the estimate of β, P Si represents the true value of stray loss, ρ τ is a weight coefficient, which is usually defined as a piecewise linear function, and the mathematical description is as follows:

[0017]

[0018] ρ(r) is a target function about residual r, which is used to define the contribution size of each residual to the target function value;

[0019] Joint formula (2), (3), the following formula can be obtained:

[0020]

[0021] b4) select the target function ρ(r)

[0022] The target function ρ(r) requires the following properties:

[0023] (1) always non-negative, that is, ρ(r) ≥ 0;

[0024] (2) when the residual is zero, the function value is also zero;

[0025] (3) meet the symmetry;

[0026] (4) meet the monotonicity;

[0027] b5) according to the selected target function, joint formula (1), (4), the numerical value of β1(τ) and β0(τ) is obtained by interior point method or analytical method, and the equation

[0028] According to the attribute requirement of the objective function p(r), the objective function p(r) recommends selection of p(r)=r 2 , Huber function or bisquare function, or other functions with the above attributes.

[0029] Compared with the prior art, the present application has the following beneficial effects:

[0030] Under the previous standard, the calculation of the stray loss is only through the least square method and through 6 groups of sample data to obtain the stray loss equation In this case, the obtained equation is too large, and at the same time, when facing an abnormal point, it is extremely susceptible to the influence of the abnormal point, and in the case of small data quantity, the error is more biased. The present application starts from the quantile regression method, and in the process of quantile regression analysis, the properties of p(r) (objective function about residual r) are defined, so that people can choose different objective functions for regression analysis. In addition, in the process of quantile regression, the present application obtains different curves by selecting different quantiles, selects the curve that best meets the fitting standard through the test of goodness of fit, and can reduce the influence of abnormal points on the curve. DETAILED DESCRIPTION

[0031] First step, establish the quantile regression model of the dependent variable stray loss

[0032] In the national standard requirement, the stray loss Ps satisfies the following equation:

[0033] P S = AT 2 +B (1-1)

[0034] T 2 is the shaft torque, A and B are coefficients;

[0035] In the quantile fitting theory, the dependent variable is usually modeled as a linear model containing quantiles, as follows:

[0036]

[0037] Where p is the number of dependent variables, r is the residual, and β is the parameter vector, which can be estimated by the independent variable x and the quantile τ.

[0038] According to formula (1-1), the stray loss P S has a linear relationship with T 2 , and according to the quantile fitting theory, the following quantile regression model can be directly established:

[0039]

[0040] T is the estimated value of stray loss. 2 Let τ represent shaft torque, τ represent quantiles, and β be a collective term for β1 and β0, representing a parameter vector. Specifically, β1 and β0 represent T, T, and β0, respectively. 2 The coefficients and residuals, at a fixed quantile, correspond to respectively A and B.

[0041] The quantile regression model in this embodiment is based on the stray loss P in the national standard. S With T 2 The univariate linear relationship is established. If the theory is extended in the future, P S If the model has a linear relationship with multiple independent variables, then this invention can establish a multiple regression model, that is, a model with more than one independent variable.

[0042] The second step involves setting quantile values ​​and minimizing the residual between the true and estimated stray loss values ​​to estimate the parameter vectors corresponding to different quantiles, thus obtaining a set of fitting curves for the stray loss sample data.

[0043] In this embodiment, the parameter vector β is estimated by minimizing an objective function ρ(r) with respect to the residual r. P Si This represents the true value of stray loss. This represents the stray loss estimate, and the ρ(r) function is used to define the contribution of each residual to the objective function value.

[0044] Generally speaking, the ρ(r) function is required to have the following properties: (1) it is always non-negative, that is, ρ(r)≥0; (2) when the error is zero, its function value is also zero, that is, ρ(r)=0; (3) it satisfies symmetry; (4) it satisfies monotonicity. Generally, ρ(r)=r is chosen. 2 Alternatively, the Huber function, bisquare function, or other functions with the aforementioned properties can be selected. Different functions will have different effects on curve fitting. In this embodiment, ρ(r) = r is used. 2 .

[0045] The estimation process for the parameter vector β is as follows:

[0046] Given N observation samples (the observation samples are the obtained data, i.e., N x-coordinates T), 2 The vertical axis is The parameter vector β is evaluated using the following formula:

[0047]

[0048] in It is an estimate of β; ρ τwhere w is the weight coefficient, which is usually defined as a piecewise linear function, mathematically described as follows:

[0049]

[0050] Let ρ(r) = r 2 Substituting, the function ρ τ (r) is written as:

[0051]

[0052] If the quantile τ corresponds to the quantile point (T m 2 , P Sm ) (indicating that the proportion of points less than the abscissa T m 2 is τ), then the formulas (1-4) and (1-6) can be converted into the following formula:

[0053]

[0054] Substituting (i.e. formula (1-3)) into formula (1-7), the parameter vector β corresponding to different quantiles τ can be solved.

[0055] The equation of the embodiment belongs to a linear programming problem, and therefore it is recommended to use the analytic method and the interior point method to optimize and solve, to obtain the linear equation of .

[0056] τ takes a value between 0 and 1, and is usually taken at a certain interval.

[0057] Third step, performing goodness-of-fit test on the curve obtained in the second step, screening out the curve with the best fitting effect, to determine the motor efficiency stray loss

[0058] The goodness-of-fit test formula is:

[0059]

[0060] Where R 1 (τ) is the goodness-of-fit, and the closer the value is to 1, the better the fitting effect.

[0061]

[0062] Finally, the patent can analyze and explain the influence of the explanatory variable on the explanatory variable at a specific quantile through different quantiles.

[0063] Under the previous standard, the calculation of stray loss is only through the least square method and through 6 groups of sample data to obtain the stray loss equation In this case, the equation deviation is too large, and when facing abnormal points, it is easily affected by abnormal points, and in the case of small data, the error is more biased. The present technology starts from quantile regression method, and in the process of quantile regression analysis, the properties of p(r) (objective function about residual r) are defined, so that people can choose different objective functions for regression analysis. In addition, in the process of quantile regression, different curves are obtained by selecting different quantiles, and the curve that best meets the fitting standard is selected by goodness-of-fit test, so that the influence of abnormal points on the curve is reduced.

[0064] In addition, it should be noted that a plurality of curves of different quantiles can also be analyzed locally. By segmenting a curve, the goodness-of-fit of different segments is tested, and then the different segments of different curves are analyzed and tested to obtain the best curve segment in each interval and the corresponding quantile of the segment.

Claims

1. A method for calculating stray losses in motor efficiency, characterized in that, Includes the following steps: a) Establish a quantile regression model with stray loss as the dependent variable: T is the estimated value of stray loss. 2 Let τ represent shaft torque, β represent quantiles, and β1(τ) and β0(τ) be a collective term for the parameter vector. Specifically, β1 and β0 represent the torque T at quantile τ. 2 The coefficients and residuals; b) Set quantile values ​​and estimate the parameter vectors corresponding to different quantiles by minimizing the residual between the true and estimated values ​​of stray loss, thus obtaining a set of fitting curves for the stray loss sample data; c) Perform a goodness-of-fit test on the curve in step b), and select the curve with the best fit to determine the stray loss of motor efficiency. Alternatively, the curve in step b) can be segmented, and the goodness-of-fit test can be performed on each segment to obtain the curve segment with the best fitting effect in each interval, and the motor efficiency stray loss can be determined accordingly. In step b), the fitting curve for the stray loss sample data is obtained as follows: b1) Determine the value of the quantile τ; b2) Find the quantile corresponding to quantile τ: (T m 2 P Sm ); b3) For a given set of N observation samples, the parameter vector β is evaluated using the following formula: in, It is an estimate of β. P Si ρ represents the true value of stray loss. τ The weighting coefficients are typically defined as a piecewise linear function, mathematically described as follows: ρ(r) is an objective function with respect to the residual r, used to define the contribution of each residual to the objective function value; Combining formulas (2) and (3), we can obtain the following equation: b4) Choose the objective function ρ(r) The objective function ρ(r) is required to have the following properties: (1) It is always non-negative, that is, ρ(r)≥0; (2) When the residual is zero, its function value is also zero; (3) It satisfies symmetry; (4) It satisfies monotonicity; b5) Based on the selected objective function, and combining formulas (1) and (4), the values ​​of β1(τ) and β0(τ) are obtained through the interior point method or analytical method, thus obtaining the equation.

2. The method for calculating stray losses in motor efficiency according to claim 1, characterized in that, The objective function ρ(r) is chosen to be ρ(r) = r. 2 The Huber function or the bisquare function.

Images