Markov Chain based Very Short Term Load Forecasting realizing Conditional Expectation

Abstract:
The short-cycle time resolved energy consumption data obtained from smart metering devices allow the energy consumption profile of households to be determined and forecasted using methods of predictive analytics. A Markov chain based method realizing conditional expectation was formulated using the conditional transition probability between consecutive states. The direct formulation as the conditional expectation is a new form of evaluating a Markov chain based model for an approach to forecasting. It was found to give comparable results to those from autoregressive methods. In addition, a key advantage of the Markov chain based method of conditional expectation is its simplicity, computational speed and stability. It comprises the forecast of the conditional and the unconditional probability distribution.