An improved ant colony clustering for power load forecasting problem

Ant colony algorithms have been recently suggested for short-term electric load forecasting by a large number of researchers. As we know that the forecasting accuracy is influenced by the distributed feature of load sample space, and the complex nonlinear relation, which is formed by the sensibility of external weather factors to power load, will also reduce the accuracy of forecasting. In this paper, an improved ant colony clustering (IACC) is put forward to raise the accuracy of electric load forecasting. The merits of IACC were parallel search optimum and the dynamic method to adjust the parameter of evaporation coefficient. What's more, this proposed algorithm enhanced the heuristic function to accelerate the searching process. Then, the performance of IACC in actual load system has shown its superiority, that is, its sensitivity and resolution to climatic anomaly circumstances, high temperature, festival and holiday condition is higher than ant colony optimization algorithm (ACOA). In addition, IACC is more exquisite and even of the clustering characteristics on the similarity of load curve profile. The IACC clustering analysis has shown that it has a most important significance to improve the accuracy of STLF.