Segmented power demand forecasting using stochastic model

This paper is based on the research of a new method to forecast power demand. Conventionally, researchers have extracted the implicit periodicity contained in the long-term power demand curve. In contrast, we have divided the long-term power demand curve into a large number of small segments and treated the power demand in each segment as a random variable without memory. Thus, by using the stochastic model and the adaptive mechanism, we can now update the relevant statistic parameters to provide an accurate forecasting result. Simulations run using this method provide an accurate approximation of the real power demand, requiring only a small amount of real-time information. This effectively reduces the operational overheads of smart grid-based power demand forecasting. The only negative effect is a negligible loss in accuracy, well under the acceptable standards. Admittedly, it is not the most accurate method to forecasting power demand, but its low system complexity outperforms over other methods.