Energy Efficiency Improvement of Industrial Enterprise Based on Machine Learning Electricity Tariff Forecasting

Abstract

One of the most important tasks for a large industrial enterprise is to reduce the production cost. The electricity cost for each industrial enterprise is formed by many factors, some of which can be influenced to reduce final costs. An overview of existing methods of load regulation is presented, as well as an assessment of their feasibility and efficiency in terms of reducing the cost of electricity and power. Mid-term forecasting electricity tariff rate to change the load schedule can reduce the company's electricity costs. The possibility of building a machine learning model predicting the retail market hourly electricity tariff rate has been studied. Modeling was performed on the publicly available data of electricity prices in the Novosibirsk region (Russia) for 2018–2020. It was found out that Extreme Gradient Boosting over regression decision trees can predict the hourly electricity tariff rate for a month ahead with a mean absolute percentage error of 4 %.