Ant colony algorithm (ACA), inspired by the foodsearching behavior of ants, is an evolutionary algorithm and performs well in discrete optimization. Ant colony algorithms have been recently suggested for short-term load forecasting (STLF) by a large number of researchers. In this paper, an Improved Ant Colony Clustering (IACC) based on Ant Colony Algorithm was put forward. In IACC, each load data was represented by an ant, making use of the parallel optimization characteristics of ant colony algorithm and the ability of volatile quotient method to adaptively change the amount of information, with improvements have been made by changing the pheromone concentration on every path and enhancing the heuristic function to accelerate the searching process. Experiments and comparisons are done to show that the IACC is an efficient and effective approach, not only IACC increased the STLF accuracy, but also IACC is more exquisite to the similarity of load curve profile. Key words: Power Load Forecasting, IACC, Colony Clustering, Load Curves