变压器的运行寿命与变压器绝缘性能直接相关。对于特高压换流变压器来说,油温预测可作为其绝缘性能评估的重要依据。为提高换流变油温预测精度,提出一种基于长短期记忆网络(long-short term memory network,LSTM)、自注意力机制(self-at...变压器的运行寿命与变压器绝缘性能直接相关。对于特高压换流变压器来说,油温预测可作为其绝缘性能评估的重要依据。为提高换流变油温预测精度,提出一种基于长短期记忆网络(long-short term memory network,LSTM)、自注意力机制(self-attention mechanism,SA)和门控循环单元(gated recurrent unit,GRU)串并行混合模型的换流变顶层油温预测方法。首先,对换流变顶层油温数据进行滚动滑窗预处理;然后,建立LSTM与SA并行的预测模型,并利用GRU对并行预测的结果进行融合,经全连接层调节后输出最终预测结果。对比实验表明,相较于单一预测模型,采用混合预测模型在换流变顶层油温预测中可以取得更高的精度。展开更多
基金supported by National High-tech Research and Development Program of China(863 Program)(2009AA04Z416) National Science Foundation of China(51021005) Scientific Innovation of Colleges and Universities(Project v-200704)
基金supported by National High-tech Research and Development Program of China(863 Program)(2009AA04Z416) National Science Foundation of China(51021005) Scientific Innovation of Colleges and Universities(200704)
基金Supported by National Natural Science Foundation of China (NSFC)(50438070)Program for New Century Excellent Talents in University of China (NCET-04-0095)Ph.D. Programs Foundation of Ministry of Education of China (200800030040)~~