基于深度学习的电力设备智能运行方式研究

Research on intelligent operation mode of power equipment based on deep learning

在电力系统中对用户行为数据的分析过程存在主观性强的问题,且在复杂电网环境下难以做出准确的判断。对此,文中提出一种基于改进K-means聚类算法和深度学习相结合的分析算法。在改进K-means聚类算法的基础上构造出用户行为数据分析模型,实现了对行为数据的自适应学习。通过HRF-TCN预测模型筛选出用户行为数据的关键特征,降低数据维度后利用时间卷积网络进行预测,并将预测结果传递给电力系统中的设备,使其智能化地调整运行方式。基于公开用电数据集对算法进行的可行性验证结果表明,所提算法能够对用户的行为数据进行准确分析,数据分类效果指标DBI、SC分别为0.826 4和0.440 1,预测指标Ac和F1分别为96.8%、0.967 8,与同类算法相比,其分类效果更好、精度更高。

There is a strong subjectivity issue in the analysis process of user behavior data in the power system,and it is difficult to make accurate judgments in complex power grid environments.In this regard,the article proposes an analysis algorithm based on a combination of improved K-means clustering algorithm and deep learning method.On the basis of improving the K-means clustering algorithm,a user behavior data analysis model was constructed to achieve adaptive learning of behavior data.Filter out key features of user behavior data through the HRF-TCN prediction model,reduce data dimensions,use time convolutional networks for prediction,and transmit the prediction results to devices in the power system to make intelligent operation mode adjustments.The feasibility of the proposed algorithm was verified based on a publicly available electricity dataset,and the results showed that the proposed analysis algorithm can accurately analyze user behavior data.The data classification performance indicators DBI and SC are 0.8264 and 0.4401,and the prediction indicators Ac and F1 are 96.8%and 0.9678.Compared with similar algorithms,its classification performance is better and its accuracy is higher.