«上一篇
 文章快速检索 高级检索

 应用科技  2018, Vol. 45 Issue (2): 43-48  DOI: 10.11991/yykj.201704005 0

### 引用本文

REN Reng, YUAN Haiwen. Method for evaluating health status of distribution equipment based on comprehensive discussion[J]. Applied Science and Technology, 2018, 45(2), 43-48. DOI: 10.11991/yykj.201704005.

### 文章历史

Method for evaluating health status of distribution equipment based on comprehensive discussion
REN Reng, YUAN Haiwen
School of Automation Science and Electrical Engineering, Beihang University, Beijing 100191, China
Abstract: In the process of health evaluation of distribution equipment, the assessment accuracy can be improved by introducing expert experiences. However, at present, our country is faced with the shortage of skilled labor. Following the retirement of experts, the expert experience can not be repeated and inherited. Therefore, according to the idea of people-oriented, man-machine combination and from qualitative to quantitative, this paper put forward a comprehensive research method on the evaluation of the health status of distribution equipment. In the method, the analytic hierarchy process(AHP) is used to select experts in different stages of the life cycle of power distribution equipment, weights are distributed to the power distribution equipment health index model put forward by experts, a comprehensive & integrated analysis is completed for the discussion results of all experts, the final result is obtained and a feedback is given. It is expected that the method can be used to quantify the expert experience, calculate the health index of the equipment and finally obtain the health status of power distribution equipment.
Key words: expert experience    comprehensive discussion    distribution equipment    health status    analytic hierarchy process    health index    human-oriented    human-machine integration

1 综合研讨评价方法

1.1 健康指数定义

1.2 整体研讨步骤

2 配电设备健康状态评价 2.1 选取专家

1)根据生命周期4个阶段两两比较的重要度不同构造判断矩阵A，构造矩阵根据每一次研讨设备的已使用年限情况不同而侧重点有所不同，按时间具体排序表1所示。

 ${{A}} = \left( {\begin{array}{*{20}{c}} {{a_{11}}}&{{a_{12}}}&{...}&{{a_{14}}} \\ [4pt] {{a_{21}}}&{{a_{22}}}&{...}&{{a_{24}}} \\ [4pt] \vdots &{...}&{...}& \vdots \\ [4pt] {{a_{41}}}&{{a_{42}}}&{...}&{{a_{44}}} \end{array}} \right) = {({a_{ij}})_{4 \times 4}}$

 ${a_{ij}} = \frac{{{w_i}}}{{{w_i} + {w_j}}}\left( {i = 1,2, \cdots ,5;j = 1,2, \cdots ,5} \right)$

 $\left\{ {\begin{array}{*{20}{c}}{{y_k} = {{A}}{\mu _{k - 1}}}\\{{\mu _k} = m({y_k})}\\{{\mu _k} = {y_k}/{\mu _k}}\end{array}} \right.(k = 1,2,3, \cdots )$ (1)

3)最后得到专家的排序向量为

 ${{{A}}_d}{{{W}}_c} = \left( {\begin{array}{*{20}{c}} { y_{11} }&{ y_{21} }&{...}&{ y_{41} } \\ { y_{12} }&{ y_{22} }&{...}&{ y_{42} } \\ \vdots & \cdots & \cdots & \vdots \\ { y_{1m} }&{ y_{2m} }&{...}&{ y_{4m} } \end{array}} \right).\left( {\begin{array}{*{20}{c}} {{w_1}} \\ {{w_2}} \\ {...} \\ [3pt] {{w_4}} \end{array}} \right) = \left( {\begin{array}{*{20}{c}} {{b_1}} \\ {{b_2}} \\ {...} \\ {{b_m}} \end{array}} \right) = {{B}}$

B中元素按从大到小排序即可选取与会的专家。

2.2 配置权重 2.2.1 配置权重过程

2.2.2 根据配电设备涉及到的各关键特征量的重要性对模型进行整合

1)对每个专家提出的关于配电设备健康指数评价的模型，根据专家选取的特征量的重要程度进行权值的修正。因为关于配电设备的特征量太多，以变压器为例，通过绛维的方式从电气性能、理化性能和自然因素3个维度选取关键特征量，选取参考标准如表3。在这个过程中根据特征量对最后健康指数评价的影响程度的不同予以不同的标记值。特征量类值 ${x_i}$ 为专家选用的这类特征值的标记值的总和，最后计算关于专家t的特征量的权值影响度

 ${e_i} = \sum {{x_i}}$

2)根据配电设备评价领域的影响力、对设备健康状态熟悉程度、在岗年限3项专家信息对专家进行标记。标记结果记录如表4。在这3项专家信息中，对设备健康状态熟悉程度尤为重要，所以它的分值划分为比去其他几项略高。现假设有m个专家，则将专家t定量化后的权值系数为

 ${w_t} = {a_t} + {b_t} + {c_t}$

 $\lambda = \frac{{{w_t} + {e_t}}}{{\sum\limits_{t = 1}^m {({w_t} + {e_t})} }}$

3）根据以往专家研讨的准确率采用基于历史信息修正的智能算法修正权重。修正权重主要依靠差值、方差以及与准确度之间的方差3个参数进行衡量。设标记值的准确值为1，通过专家给出的健康指数与设备最终所得的健康指数进行比较，具体比较内容如表5所示。

 ${\bar S^2} = \frac{{\sum\limits_{i = 1}^n {{{({x_i} - 1)}^2}} }}{{n - 1}}$

 $\bar \alpha = \frac{{\sum\limits_{i = 0}^n {{\alpha _i}} }}{n}$

 $\bar \beta = \frac{{\sum\limits_{i = 0}^n {{\beta _i}} }}{n}$

 $\bar \gamma = \frac{{\sum\limits_{i = 0}^n {{\gamma _i}} }}{n}$

 $\overline {{\lambda _t}} = {\overline \alpha ^3}{\overline \beta ^2}\overline \gamma$

4）归一化处理后得到每个专家修正后的权重。

 $\lambda _t = \frac{{( w_t + e_t ) \times {{\bar \lambda }_t}}}{{\sum\limits_{t = 1}^m {( w_t + e_t ) \times {{\bar \lambda }_t}} }}$
2.3 综合健康状态结果

 ${G_I} = \sum\limits_{i = 1}^{m - 1} {\sum\limits_{j = i + 1}^m {\frac{{2{ \theta _{ij}}^2}}{{m(m - 1)}}} }$

 $\sum\limits_{t = 1}^m { A_t } \times \lambda _t$

3 实例结果与分析