Development and Optimization of Methods for Objective Interpretation of FRA Test Results

在德国使用的许多变压器正接近其预计使用寿命的终点。为了确保尽可能长时间的安全运行，防止发生故障，必须对绝缘系统的状况进行监测。传递函数的评定是确定变压器绕组机械状态的一种比较方法。传递函数评估采用了众所周知的频响分析（FRA）技术。以往的研究已经证明，联邦铁路网勘测局可以提供可靠的信息，了解电力变压器内部有源部件的机械完整性，而无需拆卸机组。他们产生了一种标准化的测量程序，使测量结果相互比较。但是，需要在测量后决定变压器是否完好无损，是否需要在进一步通电前进行维修。这一步被称为联邦铁路局解释。然而，人类对测量结果的直接解释是主观的，因此可能容易出错。当前项目的主要目标是发展、改进和优化对森林资源评估结果进行客观解释的方法。不同的FRA解释方法将在电力变压器中进行检验。为了测试不同的解释技术，一组具有参考痕迹（指纹）和已知机械缺陷痕迹的FRA数据是必要的。典型机械故障模式的FRA数据应通过高频变压器建模生成数据，逐步实现模型绕组的不同机械缺陷，并从不同的公用事业公司、诊断公司和工作组收集现场的真实案例数据来收集。对这些数据集应采用不同的数值指标，并相互比较，以评估其适用性和规范解释程序的能力。作为最后一步，应进行一项研究，以审查为选定的指数确定阈值的可能性。此外，基于人工智能的解释技术将应用于数据集，以确定不同断层的类型和程度。

Many of the transformers used in Germany are approaching the end of their projected service life. To ensure safe operation as long as possible and to prevent a failure, the condition of the insulation system has to be monitored. The evaluation of the transfer function is a comparative method for determining the mechanical state of the transformer windings. Transfer function assessment is carried out using the well-known frequency response analysis (FRA) technique. Previous studies have proved that the FRA can provide reliable information about the mechanical integrity of active parts inside the power transformer without any needs to dismantle the unit. They resulted in a standardized measuring procedure to make the measurements comparable with each other. However, one needs to make a decision about the transformer after the measurement, whether it is intact or needs to be repaired before further energization. This step is known as the FRA interpretation.However, the direct interpretation of the measurement results by a human being is subjective and therefore potentially prone to errors. The main goal of the current project is to develop, improve and optimize methods for objective interpretation of the FRA results. Different FRA interpretation methods are to be examined in power transformers. In order to test different interpretation techniques, a data set of FRA data with reference traces (fingerprint) and traces of known mechanical defects is necessary. FRA data for typical mechanical failure modes should be collected by means of data generation by high frequency transformer modeling, stepwise implementation of different mechanical defects in model windings and collection of real case data from the field from different utilities, diagnosis companies and working groups. Different numerical indices should be applied to these data sets and compared with each other in order to assess their applicability and ability to standardize the procedure of interpretation. As the final step, a study should be carried out to examine the possibility of defining thresholds for the selected indices. Additionally interpretation techniques based on artificial intelligence will be applied to the data sets in order to determine the type and extent of different faults.