再论模式识别法在油中放电超声定位中的研究

Re-Study on Pattern Recognition in Acoustic Location of Partial Discharge in Oil

在电力变压器的局部放电在线检测中 ,放电点定位是一个非常重要的环节。由于在检测过程中各种干扰的存在 ,通常所得到的放电点到达超声传感器的时间或时延和实际结果有一定的误差 ,所以 ,定位算法是决定定位结果是否成功的关键。针对模式识别算法在变压器超声定位中存在的某些缺陷 ,提出了新的改进算法 ,在模式识别的逼近过程中增加了扰动这个中间环节作为跳出局部最优的手段。算例表明 ,和传统的模式识别法相比 ,改进算法除了能够大大提高运算速度以外 ,还能够有效地取得全局最优点。

For online detection of partial discharge of power transformer,to locate the discharging point is very important.Becarse of interference existing in detection process,the time delay from discharging point to acoustic sensor differs from the real result.The location algorithm greatly influences the successful measurement.To overcome some defects of pattern recognition in the acoustic location of partial discharge of transformer,a new location algorithm is proposed.It applies disturbance as a means to jump out from partial optimization in convergence process.The calculation example shows that the improved algorithm increases calculation speed and effectively achieves global optimization.