Early warning method for power supply service quality based on three-way decision theory and LSTM neural network

Abstract

An efficient early-warning method for power supply service quality(PSSQ) is of great significance to optimize the customer experience of power service and ensure the security of power systems. Based on the power service data from the customer side, this paper proposes a early-warning method for PSSQ, which is based on three-way decision theory and long–short term memory (LSTM) network. First of all, four early-warning indicators (i.e., the proportion of complaints work orders indicator, the proportion of responsible work orders indicator, the proportion of duplicate work orders indicator and the average processing time indicator) of PSSQ are proposed according to requirements from the customer side. Secondly, the early-warning value of PSSQ is determined based on LSTM neural network and historical data. Then, the threshold of early warning decisions is determined based on the three-way decision theory. Finally, three local power supply companies in Zhejiang province are taken for case study to prove the effectiveness of the PSSQ early-warning method proposed.