DESIGN OF AN INTELLIGENT INDUSTRIAL ENERGY MANAGEMENT SYSTEM BASED ON ARTIFICIAL INTELLIGENCE AND INTERNET OF THINGS

Abstract

Industrial electricity tariffs in most jurisdictions combine a consumption charge with a demand charge levied on the highest metered load within a billing period, yet the peak-shaving strategies commonly deployed to manage this second charge are almost universally reactive: a threshold is set on total facility demand, and flexible loads are shed once that threshold is crossed, without reference to which loads can be curtailed at least production cost or to how demand is likely to evolve over the remainder of the billing period. This study proposes Predictive Load-Shaping and Equipment Coordination (PLSEC), a framework in which a probabilistic short-horizon demand forecast is generated from historical consumption, production-schedule, and weather data, and a model-predictive-control dispatch layer uses this forecast to pre-emptively and selectively adjust the operation of flexible industrial loads — compressed-air banks, chillers, and batch thermal processes — so that the facility's realized demand trajectory approaches its billing-period peak only when doing so is unavoidable given production commitments, rather than being shed reactively after a fixed threshold is breached. The framework was evaluated on a representative multi-process industrial facility through comparative modeling of three regimes: a fixed-schedule baseline, existing threshold-based reactive peak shaving, and the proposed PLSEC system. The proposed system reduced monthly peak demand by 28%, cut equipment cycling nearly in half relative to the reactive approach, and lowered total monthly electricity cost by 19% relative to baseline, with a five-year net present value more than double that of the standard reactive approach.