Prediction Of Power And Load With Uncertainty And Economic Dispatch Of Microgrid

The economic dispatch of distributed generations for the microgrid is great important for guaranteeing the reliable power supply and economic operations,and has been becoming a hot topic in the research of microgrid.However,such a dispatch is quite different from traditional grid due to the intermittent,randomness and fluctuations of the distributed generations and loads in the microgrid system.One feasible and effective solution for this kind of dispatch is to carry out more precise predictions on the uncertain renewable generations and loads,and then perform the dispatch with such uncertainties.To this end,we here emphasize on the following researches on uncertain predictions and dispatch of microgrid.1)Interval prediction of wind power with uncertainty based on echo state network: A quantified method for addressing the essential uncertainty of wind power forecasting is proposed by using echo state network,which is aiming at the influences of uncertainties of the wind power prediction in the traditional methods only using point samples.In the proposed method,a new interval samples is first proposed to reflect the uncertainties of wind power data based on similar days and similar intervals method;then,time series interval prediction of wind power based on echo state network is given;finally,the simulations validate the effectiveness of the proposed method by analyzing the prediction interval coverage probability(PICP)and mean prediction interval width(MPIW).2)Interval prediction of microgrid loads based on ensemble extreme learning machine: The great fluctuations of loads in microgrid are first illustrated by analyzing varied loads data,which indicates that typical global prediction is quite difficult to ensure the accuracy of prediction and reflect uncertainties;then,the strategy for selecting critical factors of load forecasting is presented,and the criterion of similar days is followed given to select appropriate samples based on a new time piece-wised correlation and date type-temperature correlation.The ensemble extreme learning machines are constructed to forecast loads based on the selected loads data of similar days,and the interval uncertainties are obtained by using Bootstrap sampling method.The simulation results demonstrate the effectiveness of the proposed method.3)Multi-objective economic dispatch of microgrid with uncertainty based on NSGA-II: The multi-objective economic dispatch model of microgrid which contains wind power and other distributed generations is first constructed based on researches of chapter 3 and 4,and the real-time energy storage capacity,spot pricing and other factors are also considered.The time piece-wised optimization strategy is presented according to the interval predictions of wind power and loads with uncertainty.The solutions obtained with NSGA-II are presented when the wind power does not meet the requirement of microgrid loads.The simulation results further analysis the influence on the dispatch with the fluctuations of wind power and loads,which illustrates the feasibility of our studies.The prediction methods proposed in this paper can better reflect the possible ranges of wind power and loads.The dispatch results considering uncertainties are more superior to traditional methods in adaptability and robust.Our research is expected to provide a more reliable theoretical reference for the practical applications.