Neural Network Model with Time Series for the Prediction of the Electric Field in the East Lima Zone, Peru

Abstract

Global warming and climate change is a latent problem nowadays because it affects the quality of life of living beings that inhabit an electric planet; therefore, the atmosphere is charged with ions that constantly interact and achieve a continuous balance. Likewise, when the determined value of the electric field is exceeded in one location, this produces an electric discharge, which varies with the time of the day, month and seasons. The variation of the electric field in the troposphere of the campus of the Universidad Peruana Unión, located in the area of East Lima, Peru, has been evaluated using a EFM-100 Sensor equipment which measures the electric field during the seasons of the year, and this study aims to predict the future measurements using artificial intelligence. The area of East Lima was mapped and the EFM-100 sensor was set for its exact location and to report outputs of the electric field within a radius of 37 km. A neural network model was found that was supported by the descending gradient algorithm and the Levenberg-Marquardt algorithm of the MatLab libraries in the 2018 version. The neural network model had a mean square error (MSE) of 0.476184, the validation was 0.558515 and the testing was 0.464005. Finally, an electric field of 0.1682 v/m was obtained in the summer season, −0.66 v/m in autumn, −1.62 v/m in winter and −1.43 v/m in spring.