Research On Online Intelligent Transformer Condition Monitoring System

As one of the most important primary equipments in the electricity network, the transformer’s stability and reliablity are very important to industrial production and people’s daily life. Tranditional method of transformer maintenace is power outage maintenace, this brings too much unnecessary cost, lack of power supply during maintenance, faults unpredictable during un-maintenance period. To solve these problems, research institutes around the world put forward online transformer condition monitoring systems. But most of the monitoring systems is for a single gas monitor, that lead to fuzzy diagnosis, and the fault type is too coarse, can not achieve efficient and accurate fault diagnosis.In this case, a multi-parameter online monitoring system is established by this paper.To solve the non-linear relationship between status and faults of the transformer, a BP neural network aims to achieve IEC algorithm is introduced, and the procedure to establish a BP neural network is discussed in detail. Then some improved algorithm that can remedy the disadvantages of the traditional neural network, like lost in locality minimum point and the convergence rate is too slow, are discussed and simulated, the appropriate algorithm is selected in line with the simulation result. With regard to multi-parameter monitoring system, relevant data processing method is introduced to solve the problem that the differences of absolute value between data provided by different systems are too huge.In response to the blurred diagnosis and the coarse forecast results, this paper proposes the necessary to establish a multi-parameter online monitoring system, then discusses the composition of the system in accordance with the characteristic of different systems. List the type of faults that different systems can diagnose, and the same faults that certain different systems can monitor. Then this paper creates an overlap region according to the data and proposes the establishing strategy of the comprehensive system to make real-time monitoring of the transformer.At last, based on resilient BP neural network and the overlap region, this paper put forwards an online transformer condition system to narrow the scope of the fault diagnosis and improve the accuracy of fault diagnosis. Then, offline validation test is taken to verify the effectiveness of the system. The experimental results indicate that the system is reliable and effective.