Industrial manufacturing processes involve various types of electrical load profile, which can generally be categorized into continuous-operation loads and start–stop operation loads. These load profiles are often highly fluctuating rather than stable, forcing factories to contract demand capacities significantly higher than their average power consumption, thereby increasing basic electricity charges. By installing a small-capacity battery energy storage system at the equipment front end and utilizing its fast charge–discharge characteristics to discharge during heavy-load periods and charge during light-load periods, the total power demand of the supply line can be effectively smoothed, resulting in a substantial reduction in contracted demand capacity. In this study, a sensor-based power monitoring system consisting of current sensors, power transducers, and a programmable logic controller is used to acquire high-resolution electrical load data from industrial machines. The measured sensing data were used as the input for short-term load forecasting and battery energy storage control. We focus on industrial machinery exhibiting pronounced periodic start–stop operating characteristics, in which short-term actions on the order of seconds or minutes frequently generate peak power demands that compel users to increase contracted capacity. Time-series forecasting models, including the autoregressive, autoregressive integrated moving average, and seasonal autoregressive integrated moving average models, are first applied to predict the load behavior of such equipment. On the basis of the forecasting results, a smoothing-baseline-based load smoothing control strategy is proposed to regulate the charge–discharge operation of the energy storage system, thereby stabilizing power demand and enabling an economic benefit analysis. The results verify the feasibility of applying energy storage systems to smooth the power demand fluctuations of industrial machinery and to effectively reduce basic electricity charges.