Modeling of Predictable Variations in Multi-Time Instant Ambient Temperature Time Series

Abstract

This paper effectively devised a novel approach to characterize the predictable variations in a multi-time instant ambient temperature time series. A multiple linear regression model is used to capture the annual predictable variations accurately. The clues for predictable variations upon detailed analysis of multi-time instant daily time resolution ambient temperature data led to the invention of a set of theoretical relevant deterministic regressors forming a reducing order model. A detailed result analysis has established that the proposed model is a suitable candidate for multi-time instant daily time step data and can be extended for the risk assessment of system analysis that accounts for the temperature effect. Further, probabilistic forecasting using regression-based methods can easily combat the above-limited number of theoretical relevant regressors for decent interval forecasts. The proposed model's effectiveness is analyzed using historical ambient temperature records collected from three distinct places in India.