روش جدیدی از تشخیص محل خطا برای حفظ و نگهداری خطوط انتقال

This paper presents a new methodology for monitoring, in real-time, the conditions of the insulation of an power transmission line, detecting and locating anomalies in its operation, before the supply of power is interrupted, thus allowing for preventive maintenance. This method uses the harmonic decomposition of the leakage current to analyze the condition of line insulation and employs a neural network to locate the fault. Experimental measurements were obtained to validate the simulated results.

Electric energy is one of the most important resources for the economic development of a country, as well as promoting the satisfaction and well-being of society. Thus the Electric Power Systems must ensure a high degree of reliability in the continuity of the power delivery. However, due to the increased complexity of these systems, energy demand and the interconnection of existing systems, outages of electricity tend to be more frequent. This new and challenging scenario has required a considerable improvement in equipment and devices for fault location, control and protection, to ensure a reliable and economic operation of the power system, either under normal operating conditions or in contingencies. According to [1] the contingency conditions can be of two types: faults or failures. A fault is an unexpected deviation in at least one of the characteristics, properties or parameters of the system from an acceptable, common or standard condition, i.e. a malfunction or defect. On the other hand, a failure is a permanent interruption in the capacity of a system to perform a requested task under specific operating conditions. Among the various constituent parts of an electric power system, transmission of energy is one of the most susceptible to faults. This is due to the physical dimensions of the transmission lines (TLs) and the environment in which they are installed. These characteristics hinder their maintenance and monitoring. TL faults can be caused by the occurrence of different types of phenomena, such as end of life of its components, influence of the environment, pollution, humidity or heat, and also possible accidents, such as mechanical shocks. Due to its length, the detection and location of faults in TLs is very important because it reduces the shutdown time, allowing quicker maintenance, and contributing to the return to normal operation of the system [2] and [3]. For the verification of a fault in operation of a system, it is necessary to identify parameters that indicate this fault. In the case TLs the insulation conditions determine the operating state of the TL. Since the insulation depends on the resistance and dielectric strength of the TL, an observable variable that can be used to indicate the level of insulation is the leakage current. This is because the lower the TL insulation level, the higher its leakage current. This paper proposes a method for fault location on transmission lines using harmonic decomposition of the leakage current. The main tools are the mathematical model of the transmission section to simulate an TL, and an artificial neural network (ANN) of the perceptron multilayer backpropagation type to locate faults inserted into the model. Actual measurements of voltage and current, obtained through power meters installed in substations, are used to validate the model used in location. The methodology described is registered as an international patent under number 00002200600491835 of April 19, 2006.

This work presents a methodology for detection and location faults, and provides variables to assess the condition of insulation of the TL, using the idea of harmonic decomposition of the leakage current of a TL. An ANN and a mathematical model of the line were used as tools. The validation of the mathematical model was based on the comparison of the results calculated by the model with measurements obtained from a stretch of real TL. The results shown were considered satisfactory for fault location and for the determination of capacitance values that create faults. With these two parameters it is possible to analyze the insulation conditions of a line in a particular stretch. The ANN was successfully tested with data acquired on other days. It is noteworthy that there is a very tenuous limit between a fault situation and the normal operating condition, as there are countless situations of faults and normal operation conditions. The results were considered good and applicable. Further research is needed to provide conclusive predictive maintenance. For example, it is necessary to determine whether this approach works for long lines, in which case there is a transposition of lines and the capacitances are crossed. Furthermore, there are no records available of the capacitance values for defects or normal conditions for a TL. This is a problem that makes the faults classification more difficult. The next phase of the research is the implementation of the model and the program described in this article in real-time monitoring to validate the detection and diagnosis of faults. This paper presents a real possibility of prediction and locating failures for a transmission line using a conceptually different and innovative idea.