Electricity is critical for industrial and economic advancement, as well as a driving force for sustainable development. This study collects the energy consumption data of annealing processes from an annealing furnace of a co-operating steel forging plant. We propose a CDF-based symbolic time-series data mining and analytic framework for electricity consumption analysis and prediction of machine operating states by machine-learning techniques. We computed the breakpoint value relying on a density-based notion – namely, the cumulative distribution function (CDF) – to improve the original breakpoint table in the SAX algorithm for symbolizing the time-series data. The main contribution of this work is that the modified SAX algorithm can achieve better prediction the operating state of the machine in comparison to the original SAX algorithm.