Study On Prediction Interval Estimation And Genration Scheduling Method For Power System With Large-scale Wind Power Integration

As the amount of renewable energy resources increase around the world, wind energy has gained wide spread concerns in recent years. Different from conventional generators, the large-scale integration of wind power to power system has adverse effect on stable operation and generator scheduling due to its uncertainty and variability. In tackling these issues, accurate and reliable wind power prediction becomes a meaningful tool to ensure the reliability and economical efficiency of power system. However, traditional wind power point forecast errors always exist and cannot be eliminated due to the highly volatile and uncertainy in the chaotic time series of wind power. Unlike point prediction, the prediction intervals (PIs) provide a range, within which the target will lie within with a certain probability, to estimate the potential uncertainty and risk.Therefore, by analyzing the characteristics of wind power forecast error and the original wind power time series, this paper proposes two prediction interval estimation methods. Moreover, the economic dispatch model for power system with the prediction intervals of wind power is then formulated. The main work is as follows:(1) The distribution characteristics of wind power forecast error are first studied and a novel prediction intervals estimation method based on cloud model is proposed. In this method, the cloud transformation is first used to extract the conceptual cloud of wind power forecast error series, and the probability density function of each clound model is then calculated. According to the lowest width principle of interval estimation, the PIs of wind power at a certain confidence level are finally realized. Compared to traditional forecast error distribution models, the proposed method is found to be more flexibility and effectively, which provides a reference for the study of robust generation scheduling with wind power intergration.(2) For the very short time wind power time series, unlike most existing prediction intervals construction methods, which are placed after a deterministic forecasting model with or without prior assumptions, a novel lower upper bound estimation approach based on extreme learning machine (ELM) is applied to directly construct PIs for wind power. Based on analyzing the interval forecasting error information of training dataset, a new evaluation criteria and objection function for PIs optimation is also developed in this method to obtain better PIs. The testing results demonstrate the proposed approach to be a more effective method in both prediction accuracy and computational cost, and which provides a support for the interval forecasting of very short time wind power time series.(3) By analysizing the inherent characteristics of observed wind power time series, this paper proposes a novel interval forecasting approach in which the point forecasting information is taken into consideration. In this method, the wind power time series are first decomposed into steady component and noise component using ensemble empirical mode decomposition (EEMD) and sample entropy techniques. For the steady component, the next point value is predicted by an advanced ELM method. For the unstable noise component, the noise PIs are then obtained by the ELM-based interval forecasting model. Finally, the prediction results of these components are summed to form the overall PIs. Compared to the existing methods, the proposed approach is found to be more effective in terms of higher reliability and sharpness.(4) For the solution of parameters optimization in the interval forecasting model and the economic dispatch of power system, this paper proposes a hybrid quantum bacterial foraging optimization (QBFO) algorithm. In this method, the bacteria individual is described in the quantum space and the probability density functions of each bacterial are created to update their position. Moreover, a dynamic approximation control strategy is introduced to update the chemotactic step size, which avoids the defects of the fixed swim step in bacterial foraging optimization. The experiment results on classic functions demonstrate the global convergence ability of the proposed method with better accuracy and higher probability of getting global optimum.(5) On the basis of the above research, the economic dispatch model of power system with interval forecasting information of wind power is then formulated. For the day-ahead generation scheduling, a robust unit comment model with interval forecasting values is first developed to ensure the security of power system under different scenarios. Besides, considering the characteristic components of wind power series, a hierarchical rolling scheduling model is proposed, in which both the economy and the security are taken into consideration. The simulation result demonstrates the effectivenss of propose model, and it shows to be a new way for the generation scheduling of grid integrated with wind power.