Energy consumption prediction is crucial for modern resource management and sustainable development. However, due to significant differences in the energy consumption mechanism of machine tools under different working conditions, traditional prediction methods face challenges. Based on this, a machine tool milling energy consumption prediction method based on working condition division in the entire working domain is proposed. Firstly, the milling process is analyzed and divided into three stages to study energy consumption characteristics. Secondly, considering the complexity of the cutting stage, the Spearman rank correlation coefficient method is used to screen key machining features, and the DBO-DBSCAN algorithm is used to cluster them. The clustering results are used as working condition classification. Finally, the improved Sparrow Search Algorithm-Support Vector Machine (ISSA-SVM) model is adopted for prediction by comparison with other models, and different working condition data sets are input. The effectiveness and superiority of the proposed method are demonstrated through practical examples.