基于改进深度信念网络的电力欠费预警研究

Research on Electricity Arrears Alert Based on Improved Deep Belief Network

"先使用,后付费"的营销方式导致电网公司电费回收不到位,难以支撑正常运转和获得基本效益。为了解决电力欠费对电网公司的不利影响,论文提出电力欠费预警智能预测的研究方法。将电费回收分为时间和金额两部分,结合相应的关联指标,建立参数自适应的深度信念网络,通过深度学习和训练对电力欠费情况精准预测。实验结果表明,与BP神经网络相比,深度信念网络更能准确预测出用户电费回收的未来情况,有效辅助电力企业制定用电和电费预警策略。

The kind of marketing style "Buy now,Pay later" in electric power enterprise may lead to the inefficiency tariff recovery management,which can’t achieve high efficient working and benefit. In order to avoid this problem,a new tariff recovery method is proposed. The power tariff is divided into predictions of time and value,then in combination with the index and the acquirable data,a parameter adaptive deep belif network is established. Through indepth learning and training,accurate prediction of electricity arrears is made. Finally,compared with the results of BP neural network,the deep belief network can predict more accurately and help the electric power enterprises to make strategies effectively in the electricity tariff recovery.