Power system short term load forecasting based upon a combination of Markov chain and fuzzy clustering

A method based on the combination of fuzzy clustering and Markov chain models is presented. For different types of random phenomena in time series, several functions are built up respectively. State analysis of an object is carried out using the Markov chain, while fuzzy clustering is employed to the states of samples to suit the real case. Then according to state transfer, the load change is predicted. The algorithm which is used in load forecasting reaches the global optimum, when the time series have strongly properties of randomness, the algorithm works well. Simulation results show that the error can be limited to the level of 3.5%.