基于复合分析算法的智能电表数据应用

Application of Smart Meter Data Based on Composite Analysis Algorithm

随着智能电表使用的普及化,其带来的海量数据得到了广泛关注。对这些数据的分析可获得众多有价值的信息。用户用电量预测是泛在电力物联网建设的核心内容。利用智能电表海量数据实现用户用电量的精准预测,是智能电表数据的重要应用方向之一。提出了基于模糊聚类、子群发现和长短期记忆(LSTM)神经网络相结合的复合算法,利用海量智能电表数据实现了用户用电量精准预测。首先,通过模糊聚类将区域内用户按用电量进行合理分类。然后,采用Apriori子群发现算法深度挖掘各子类中影响用电量的关联性因素,将各子类的相似日关联性因素和历史用电量作为LSTM的训练数据完成神经网络的训练。最后,实现了目标日的用电量精准预测。算例分析表明,该算法有效、可行。与其他预测方法相比,该算法剔除了非关联因素的影响,预测精度明显提高,实现了智能电表海量数据的高效利用。

With the popularization of smart meters,the massive amount of data they bring has gained widespread attention.The analysis of these data can obtain numerous valuable information.Customer electricity consumption prediction is the core of the construction of ubiquitous power Internet of Things.Using the massive data of smart meters to realize the accurate prediction of user electricity consumption is one of the important application directions of smart meter data.A composite algorithm based on the combination of fuzzy clustering,subgroup discovery and long short-term memory(LSTM)neural network is proposed to achieve accurate prediction of user electricity consumption using massive smart meter data.Firstly,the users in the region are reasonably classified by electricity consumption through fuzzy clustering.Then,the Apriori subgroup discovery algorithm is used to deeply explore the correlation factors affecting electricity consumption in each subgroup,and the similar daily correlation factors and historical electricity consumption of each subgroup are used as the training data of the LSTM to complete the training of the neural network.Finally,the accurate prediction of electricity consumption on the target day is realized.The analysis of the algorithm shows that the algorithm is effective and feasible.Compared with other prediction methods,the algorithm eliminates the influence of non-correlated factors,improves the prediction accuracy significantly,and realizes the efficient use of massive data of smart meters.