Influência de variáveis meteorológicas em sistemas de telemetria aplicados a tanques de GLP

Abstract

This Undergraduate Thesis analyzes the influence of meteorological variables on the signal strength of an Otodata TM5240 telemetry device installed in a vertical cylindrical P190 tank of Liquefied Petroleum Gas (LPG), located in the region of São Luís, Maranhão, Brazil. The study is characterized as a case study with a quantitative approach, based on the integration of historical telemetry data extracted from the Neevo/Otodata platform with official meteorological data from the Instituto Nacional de Meteorologia (INMET), referring to the period from January to December 2025.

The datasets were integrated through temporal matching, associating each telemetry record with the closest meteorological observation within a maximum tolerance of one hour, resulting in a consolidated dataset containing 788 records and 25 variables. After data preprocessing, which included the removal of irrelevant columns, variable standardization, and missing value imputation, exploratory and correlation analyses were conducted using the Pearson correlation coefficient. Subsequently, the Random Forest Regressor algorithm was applied with an explanatory focus in order to estimate the relative importance of climatic variables on the signal strength.

Additionally, the Mann–Kendall trend test was applied to the signal strength time series, indicating the presence of a statistically significant increasing trend over the analyzed period (τ = 0.4232; p < 0.001). This result demonstrates a progressive increase in the average signal level, suggesting the occurrence of structural or operational changes that positively impacted the communication link performance over time.

The Random Forest results indicated that air temperature, wind gusts and wind speed, minimum and maximum temperatures, and atmospheric pressure were the most relevant meteorological variables in explaining the observed variations in signal strength. The model achieved a mean squared error (MSE) of 108.86 and a coefficient of determination (R²) of 0.2671, indicating that although meteorological conditions exert a significant influence, a considerable portion of the signal variability is still associated with physical and operational factors not represented by the climatic variables.

It is concluded that the incorporation of meteorological information and inferential statistical analyses in the evaluation of telemetry systems can provide relevant technical support for planning monitoring actions and preventive maintenance, contributing to increased operational reliability in systems applied to LPG distribution.