Application of improved gray Markov model in power load forecasting

This paper discussed an improved gray forecasting method based on fuzzy clustering and Markov correction. Gray forecasting model is suitable for load forecasting for its simple principle and convenient calculation in small sample. But there are some limitations for great fluctuating load. In this paper, get initial forecasting load with GM(1,1) model at first. Then calculate the relative error of each time. On the basis of that, use fuzzy clustering method to classify the error and divide into some reasonable groups. At last, predict the most possible state of next error with multistep Markov transition probability matrix in order to correct the forecasting results of GM(1,1) model. The new complex algorithm can fully utilize the information of history data and further broad the application of gray forecasting method. Application results show that the error can be 1imited to the 1eve1 of 3% which is better than GM(1,1) model.