基于电气信息的变电设备状态渐变过程分析方法研究

发布时间：2018-05-26 22:29

本文选题：状态检修 + 渐变过程　；参考：《山东大学》2014年博士论文

【摘要】：变电设备的状态对电网安全可靠运行起着非常关键的作用,且随着电网规模的扩大和电气设备容量的增加,这种作用更为显著。变电设备一旦故障,将直接造成用户停电,从而带来经济损失,甚至威胁人身安全。因此,研究变电设备的潜伏故障检测、状态评估和检修措施至关重要。传统的定期检修制度因存在成本高、潜伏故障检测能力差等问题,正逐步被基于状态的检修制度替代。状态检修,即根据设备状态确定合适的检修时机和检修措施,以实现人、财、物的最优配置,其基础是设备状态的在线监测与评估。目前研究主要思路是通过综合分析电气量和非电气量监测数据,利用各种算法评估设备状态,取得了较好的实际效果。但是,状态检修更关注设备状态的变化过程,这是准确确定检修时机,进而实现设备利用率最大化、并对生产影响最小的基础。而已有研究侧重于设备当前健康状况评估,缺少对设备状态渐进变化过程的细致分析,迫切需要研究相应分析方法。利用丰富的电气信息,建立其与设备状态之间的关联,为设备状态评估提供辅助分析信息具有无需附加额外装置,量测数据丰富,获取方便的优势。因此,论文基于电气信息,从数据挖掘角度出发,对比设备端口模型参数概率分布差异实现渐变过程特征提取,分析了雷击、外部短路故障等冲击对变电设备状态变化过程的影响,并在把握渐变过程规律基础上提取未来变化趋势特征,以期为检修措施的制定提供更多辅助信息,有利于状态检修的进一步实施。论文的创新性工作如下： (1)变电设备状态渐变过程分析方法：变电设备状态渐进变化过程是由诸多微小变化累积而成,这些微小变化可以通过基于广义伏安特性构建设备端口模型的参数变化规律来间接反映。然而,外界环境和量测误差的影响,导致相应参数辨识结果呈现较强的随机性,其内在趋势特征规律难以提取。因此,提出基于统计学的变电设备状态渐变过程分析方法。首先,将变电设备运行过程分成多个时段,采用非参数核密度估计法计算各个时段内设备端口模型参数的概率密度函数,并提取参数的概率特征。然后,分析不同时段内变电设备端口模型参数概率特征的差异,从而定义了四个表征变电设备状态渐变过程的指标：端口模型参数概率密度函数最大值对应参数值Ckmax,表示该时段内端口模型参数的最大可能值；各时段内Ckmax相对第一个时段内C1max的差值,表示设备损伤随运行时间的不断积累；各时段内端口模型参数相对于第一个时段内的变化概率,表示设备不断远离初始状态；各时段内端口模型参数相对于告警状态下对应参数概率分布的变化概率,表示设备逐渐靠近告警状态。最后,利用这些指标分析端口模型参数渐变过程,得到指标序列,从而为分析变电设备状态变化趋势提供辅助分析基础。所提方法从统计的角度出发,通过大量历史样本数据挖掘概率特征分析渐变过程,受少数不良数据影响小,具有良好的抗干扰能力和鲁棒性。其中,概率密度函数的计算采用非参数核密度估计法,不需要预先假设设备端口模型参数的分布,减少了主观因素的影响；变电设备端口模型参数通过偏最小二乘回归辨识得到,保证了结果准确可靠。以分析变压器绕组形变累积效应为例,通过蒙特卡洛法获取漏电感参数,实现变压器绕组形变累积过程的模拟；利用定义的四个指标对该渐变过程进行分析,结果表明该方法有效可行。 (2)冲击对变电设备状态渐变过程影响的分析方法：变电设备运行过程中,不可避免的遭受来自雷击、外部短路故障等冲击的影响,冲击导致的变电设备状态变化隐含着设备安全信息,必须引起足够重视。量化分析外部冲击带来端口模型参数的变化对于后续变化过程特征提取十分必要。但是,外界环境和量测误差造成的端口模型参数随机波动,增加了检测的难度。因此,考虑端口模型参数的随机波动特性,提出分别基于概率密度函数差异和自适应积分算法的两种检测与分析方法。前一种方法中,端口模型参数变化的检测通过分析相邻时间窗口内参数的概率分布差异实现,变化的幅度通过概率密度函数最大值对应参数值的差值反映,该方法检测准确,计算量较大,适用于冲击过后量化分析端口模型参数的变化；后一种方法中,利用相邻时间窗口内端口模型参数差值样本的均值不同对变电设备状态变化进行检测,并直接用该均值反映端口模型参数变化幅度,该方法计算快速,能及时检测端口模型参数在没有达到报警或保护动作条件时的突变。在这两种方法中,门槛值均由历史数据自适应确定,能够同时协调检测的灵敏度和准确度。通过改变变压器漏电感参数模拟雷击、短路故障等冲击造成的变压器状态变化,仿真分析结果验证了这两种方法的有效性和可靠性。 (3)间接反映设备状态的端口模型参数变化趋势特征分析方法：状态检修需要分别从长时间尺度和短时间尺度对变电设备状态的变化趋势进行把握。为此,根据提取的端口模型参数渐变过程分析指标序列,利用经验模态分解提取指标的趋势分量,建立长时间尺度下未来时段内指标的预测模型,预估达到变电设备告警状态对应端口模型参数的时段,进而为估计变电设备当前状态距离告警状态的时间进行辅助分析,为变电设备状态评估、检修措施制定提供有益的辅助依据。为详细分析未来短时间尺度下端口模型参数变化情况,提出基于状态转移概率矩阵预测概率分布的方法；通过统计相邻时间窗口内端口模型参数在各个参数变化区间的转移情况,建立状态转移概率矩阵,并预测后续时间窗口内端口模型参数的分布,进而辅助分析未来设备状态变化细节。以分析变压器绕组形变累积效应导致的变压器状态变化为例,在表征绕组形变累积过程的端口模型参数指标序列基础上,对当前参数距离告警状态对应参数值的时间进行了估计,能够为变压器状态评估提供辅助依据,有利于检修措施的制定。为模拟变压器临近告警状态的场景,利用蒙特卡洛法获取三个相邻时间窗口内漏电感参数；使用前两个时间窗口内端口模型参数样本计算状态转移概率矩阵,并预测第三个时间窗口内端口模型参数的分布；最后,计算其与直接利用蒙特卡洛模拟获得第三个窗口内参数样本的相似度,结果验证了短时间尺度下预测方法的有效性。
[Abstract]:The status of the power - changing equipment plays a very important role in the safe and reliable operation of the power grid , and with the expansion of the scale of the power grid and the increase of the capacity of the electrical equipment , the effect is more obvious . Once the substation fails , it will directly cause the user to power off , thus causing economic loss and even threatening the personal safety . Therefore , it is important to study the latent fault detection , state assessment and overhaul measures of the power transformer equipment .

In this paper , based on the analysis of electrical quantity and non - electric quantity monitoring data , it is necessary to study the state of equipment .

( 1 ) The state gradual change process analysis method of the power transformation equipment : The gradual change process of the state of the power transformation equipment is accumulated by many small changes , which can be indirectly reflected by constructing the parameter variation law of the equipment port model based on the generalized volt - ampere characteristic .
The difference between Ckmax and C1max during each time period indicates the continuous accumulation of equipment damage with running time ;
the parameter of the port model in each time period is relative to the change probability in the first time period , indicating that the device is continuously moving away from the initial state ;
in that method , a non - parametric kernel density estimation method is adopted to analyze the probability characteristic of a large number of historical sample data mining , and the influence of subjective factors is reduced by not need to pre - assume the distribution of the parameter of the equipment port model .
The parameters of the port model of the transformation equipment are identified by the partial least square regression identification , and the result is ensured to be accurate and reliable . By analyzing the deformation accumulation effect of the transformer winding , the leakage inductance parameter is obtained by Monte Carlo method , and the simulation of the deformation accumulation process of the transformer winding is realized ;
The gradient process is analyzed by using four defined indexes , and the results show that the method is effective and feasible .

( 2 ) The analysis method of the influence of the impact on the state gradual change process of the power transformation equipment : During the operation of the power transformer , it is inevitable to suffer from the impact of lightning , external short circuit fault and so on . The change of the port model parameters caused by the impact of the external environment and the measurement error is very necessary . However , the detection of the parameter change of the port model can be realized by analyzing the difference of the probability distribution of the parameters in the adjacent time windows .
in that lat method , the state change of the power transformation equipment is detected by using the mean value of the difference sample of the port model parameter in the adjacent time window , and the change amplitude of the port model parameter is directly reflected by the mean value , the method is fast , the sensitivity and the accuracy of the detection can be simultaneously coordinated by the historical data adaptive determination , and the simulation analysis result verifies the validity and the reliability of the two methods .

( 3 ) The change trend characteristic analysis method of port model parameters that indirectly reflects the state of the equipment : the state maintenance needs to grasp the change tendency of the state of the transformer equipment from the long time scale and the short time scale , respectively .
The state transition probability matrix is established by counting the transition of the port model parameters in the adjacent time windows in each parameter change interval , the distribution of the port model parameters in the subsequent time window is predicted , and the details of the state change of the future equipment are analyzed , and the time of the parameter value corresponding to the current parameter distance alarm state is estimated based on the parameter index sequence of the port model characterizing the deformation accumulation process of the transformer .
calculating the state transition probability matrix by using the port model parameter samples in the first two time windows and predicting the distribution of the port model parameters in the third time window ;
Finally , the similarity of the parameter samples in the third window is obtained by Monte Carlo simulation , and the validity of the prediction method under the short time scale is verified .
【学位授予单位】：山东大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TM711

【参考文献】