张小军, 许永新, 庄文兵, 王永强, 刘杰, 赵蓂冠. 基于水波算法−因子分析−长短时记忆网络的重要输电通道风险评估预测[J]. 现代电力, 2022, 39(3): 278-286. DOI: 10.19725/j.cnki.1007-2322.2021.0104
引用本文: 张小军, 许永新, 庄文兵, 王永强, 刘杰, 赵蓂冠. 基于水波算法−因子分析−长短时记忆网络的重要输电通道风险评估预测[J]. 现代电力, 2022, 39(3): 278-286. DOI: 10.19725/j.cnki.1007-2322.2021.0104
ZHANG Xiaojun, XU Yongxin, ZHUANG Wenbing, WANG Yongqiang, LIU Jie, ZHAO Mingguan. Risk Assessment and Prediction of Important Transmission Channel Based on Water Wave Optimization-Factor Analysis-Long and Short-Term Memory Network[J]. Modern Electric Power, 2022, 39(3): 278-286. DOI: 10.19725/j.cnki.1007-2322.2021.0104
Citation: ZHANG Xiaojun, XU Yongxin, ZHUANG Wenbing, WANG Yongqiang, LIU Jie, ZHAO Mingguan. Risk Assessment and Prediction of Important Transmission Channel Based on Water Wave Optimization-Factor Analysis-Long and Short-Term Memory Network[J]. Modern Electric Power, 2022, 39(3): 278-286. DOI: 10.19725/j.cnki.1007-2322.2021.0104

基于水波算法−因子分析−长短时记忆网络的重要输电通道风险评估预测

Risk Assessment and Prediction of Important Transmission Channel Based on Water Wave Optimization-Factor Analysis-Long and Short-Term Memory Network

  • 摘要: 重要输电通道风险评估和预测对状态检修和线路运维工作具有指导性意义,而采用传统长短时记忆(long and short-term memory,LSTM)网络对线路风险进行预测时,人为调参困难、预测精度较低,因此,提出了一种基于水波优化-因子分析-长短时记忆(water wave optimization - factor analysis - long and short-term memory,WWO-FA-LSTM)的重要输电通道风险准确评估与快速预测方法。首先,引入Levy分布、高斯–柯西变异算子和线性递减波高对WWO进行改进;其次,获取评估区多维致灾因子,并进行FA降维后作为网络输入,考虑孕灾环境敏感性和承灾体易损性计算出风险指数Rc作为网络输出;通过改进的WWO对LSTM进行不断优化,得到最优化LSTM模型;最后,采用最优化LSTM模型对重要输电通道进行风险预测。结果表明,该模型风险评估准确,模型预测较传统方法降低了误差,适用于输电通道风险评估与预测。

     

    Abstract: Risk assessment and prediction of important transmission channels are of guiding significance for condition based maintenance and line operation and maintenance. However, it is difficult to adjust parameters artificially and the prediction accuracy is low when traditional long and short-term memory (abbr. LSTM) network is used to predict line risk. For this reason, an accurate risk assessment and fast prediction method for important transmission channel based on water wave optimization-factor analysis-long and short-term memory (abbr. WWO-FA-LSTM) was proposed. Firstly, by means of leading in Levy distribution, Gauss-Cauchy mutation operator and linearly decreasing wave height the WWO was improved. Secondly, the multi-dimensional disaster causing factors in the assessment area were obtained, and after the factor analysis and dimensionality reduction these factors were regarded as network input, and considering the hazard inducing environmental sensitivity and vulnerability of disaster bearing body the risk index RC was calculated and taken as the network output. By means of improved WWO the LSTM was continuously optimized to obtain the most optimal LSTM model. Finally, by use of the most optimized LSTM model the risk prediction for important transmission channel was conducted. Results of risk prediction show that it is more accurate to assess the risk by the proposed model, and comparing with traditional methods the error from the model-based prediction is lower, so it is suitable to the risk assessment and prediction of transmission channels.

     

/

返回文章
返回