Markov-Based Stochastic Unit Commitment Considering Wind Power Forecasts

To reduce the dependence on fossil fuels and the greenhouse gas emission, the integration of wind energy has attracted worldwide attention. Stochastic unit commitment (SUC) problem with wind generation uncertainty is difficult, since wind generation is intermittent and uncertain. In the stochastic programming approach, a large number of scenarios are required to represent the stochastic wind generation, resulting in large computational effort. A Markovian approach was used to formulate the SUC problem by assuming the state of intermittent generation at a time instant summarized the information of the past in a probabilistic sense, in order to reduce computational complexity. For simplicity, wind generation state probabilities were calculated from state transition matrices, which were established based on historical data. In this paper, to improve the modeling accuracy, wind power forecasts are embedded into the Markov modeling framework, where the wind power forecast with historical forecast error is converted to discrete states with associated probabilities. In addition, corrective control actions such as load shedding and wind curtailment are considered in the Markovian approach to capture high-impact abnormal operating conditions such as sudden wind changes. Numerical testing results demonstrate the cost efficiency of the new method.