Prediction of equipment failure rates in power distribution networks based on machine-learning method

This paper explores the application of a machine learning approach to predict equipment failure rates in power distribution networks, motivated by the significant impact of power outages on citizens' daily lives and the economy. In this research, data on equipment failure rates and maintenance records were collected from power distribution networks in Baghdad, Iraq. The collected data underwent preprocessing, and features were extracted to train Adaptive Neuro-Fuzzy Inference System (ANFIS) and Periodic Autoregressive Moving Average (PARMA) time series models. To initiate the project, information regarding blackouts that occurred between January ۲۰۱۸ and December ۲۰۲۱ was retrieved from the database. The RMSE index results for the PARMA time series and ANFIS model are ۳.۵۱۸ and ۲.۲۶۴, respectively, demonstrating the superior performance of the ANFIS model in predicting equipment failure rates and its potential for future predictions. This study highlights the ANFIS model's capacity to anticipate equipment failure rates, potentially enhancing maintenance efficiency and reducing power outages in Baghdad. The error mean square was employed to evaluate the proposed model error rate.

Keywords:

Failure Rates , Machine Learning , Adaptive Neuro-Fuzzy Inference System model , Periodic Autoregressive Moving Average