Industrial Energy Forecasting Using Machine Learning and Plant Activity Metrics

This paper reports the development of a prototype AI-driven energy forecasting tool for industrial applications, designed to operate with energy metering systems and plant activity metrics. A time history of real-time plant activity data, which is typically available from SCADA-based asset activity indicators and operational parameters, is utilised along with energy timeseries data within a model regressed to correlate energy consumption with high accuracy. Prototype software was developed as a Windows Forms application and applied to an industrial case study: provision of real-time energy forecasts and optimization insights for a working sawmill are discussed. The application has been designed for future integration with existing SCADA systems, enabling seamless, automated energy monitoring, forecasting and decision support. Future enhancements of the model include incorporating specific production types and more granular operational data to further refine prediction accuracy, and application to profiling energy efficiency of process design options. The work highlights the potential for practical and low-cost AI-powered tools to enhance energy efficiency in modern manufacturing environments and process industries, including small/medium enterprises (SMEs).