Optimization of industrial energy consumption for sustainability using time-series regression and gradient descent algorithm based on historical electricity consumption data

Optimizing electricity consumption to minimize wastage and reduce cost is a major challenge in many industries. This is because, in many cases, the effect of the independent variables contributing to the total electricity consumption and cost are latent. The purpose of this study is to apply numerical techniques to identify and optimize these independent variables in order to improve sustainable energy management in industries to minimize wastage. Regression analysis was first applied to identify and decouple the independent variables to determine their individual effects on electricity consumption and cost. A cost function called the Mean Square Error (MSE) was then used to optimize these independent variables using gradient descent algorithm (GDA). In a case study, the developed approach that combines time series regression analysis with gradient descent optimization was used to analyze the electricity consumption data of an oil distribution company for the period 2015 to 2018. The results showed potential electricity savings of 124,684 kWh and cost savings of US$ 25,375 annually, when the facility is operated at optimum parameters of 0.95 power factor, 260 kVA maximum demand and 25,000 kWh active electricity consumption. The novelty of this study is that a procedure that combines time series regression analysis (RA) and gradient descent algorithm (GDA) has been developed and applied to decouple and optimize the independent variables that affect electricity consumption in an industry.