Summary: Aiming at the case-base organization in short-term load forecasting with case-based reasoning, a classifying method is presented which briefly classifies the data according to the distinguished time intervals and weekdays, and then schemes the environments by fuzzy clustering to reflect the similarity and differences among the forecasting environments. For case retrieval, a new quantitative strategy based on fuzzy sets is then proposed to improve the efficiency and govern the total procedure. A practical example shows that the proposed CBR method outperforms the normal CBR method and other competing methods in forecasting accuracy.