Forcasting Of Wind Power And Optimization For Microgrids

In recent years, microgrid technology has been developed rapidly and the study of the related technology of microgrid has become a research hot spot of smart grid technology.The construction of micro power grid has become one of the main tasks of the smart grid construction, and also become the effective ways and important auxiliary means for solving the other problems of large utility power grids, which are long construction period, high operation cost, difficult maintenance, difficult to adapt the demands of increasingly high safety, reliability, and diversification of power supply. Micro power grid has become an effective way of networking with the local resources utilized to solve the power supply in remote rural areas and islands. It is of important theoretical and realistic significance for the system security, stability, economic operation of the microgrids to reasonably predict the output power of the microsources and optimize the microgrid. In this dissertation, an in-depth research of the problems about the wind power prediction and optimization of microgrid system is made, a proposal of two kinds of wind power prediction method is performed based on the different original wind power series and the historical data, and a construction of two kinds of micro power grid optimization model of active power output is conducted according to the characteristics of the micro sources in micro power grid. The main research work and research results are as follows:1. According to the conditions with the original wind power time series being chaotic time series and the historical wind power data being large, an improved BP neural network prediction method of wind power is proposed.Firstly, the chaos phase space reconstruction technique is utilized to implement the phase space reconstruction of wind power output so as to transform one dimensional space to the multidimensional space, and it is determined whether the chaos occurs or not by computing its maximum Lyapunov exponent to obtain the best embedding dimension and delay time according to the best embedding dimension and delay time is the best structure of three layer back propagation neural network; Then the three layer back propagation neural network is constructed according to the best embedding dimension and best delay time; And then an optimization of the weights and threshold of back propagation neural network is made by using a genetic algorithm; Finally, the back propagation neural network obtained in such way is used to predict the power output on the wind farms and calculate the corresponding performance index of the predicted results.The results obtained verify the effectiveness of the method.2.Aiming at the circumstances with the original wind power time series satisfying the conditions of the empirical mode decomposition and the historical wind power data being less, a proposal of a combination forecast method of wind power is performed.First of all, a preprocessing of the original wind power is made, that is, the empirical mode decomposition technique is utilized to decompose the output power of wind power to get a series of intrinsic characteristic mode function; Then according to the characteristics of the different intrinsic characteristic mode functions, the support vector machine model optimized by wavelet neural network and cross validation parameters is employed for regression prediction; Finally all the prediction results of the subsequence are summed up as the final prediction results. Comparison of the performance index obtained by the forecast and the one by computation verifies the correctness of the method. When the conditions of both the proposed methods can be satisfied, the results predicted by the combination forecast method are even closer to the true value.3.With the renewable energy sources in microgrid deterministic and the minimum cost as the goal by considering the output in the optimal way (safe,stable, environmental and economic) in the whole system, a multi-objective optimization model of microgrid power output containing constraint conditions is established, a multi-objective optimization algorithm based on a crowded distance nondominant neighbor immune constraint is proposed,and the corresponding model is solved and optimized by the algorithm. The process of the solution of the model is as follows:First of all, the constraint conditions are transformed into an objective function so that the objective function of the model will be increased by one on the basis of the origin;Then in the individuals produced randomly, a few relatively isolated nondominant individuals are chosen as active antibodies, and according to the condition of the congestion of the active antibodies, the appropriate immune operations such as cloning, recombination is performed to strengthen the search ability of the current relatively sparse region in Pareto front. An application of the proposed multi-objective optimization algorithm in the optimization of microgrid power output is carried out both the modes of the island and connection to the utility grid.4.In order to better consider the randomness of the micro power grid, a microgrid optimization model based on probability is established, which is a kind of model with the probability contained and the minimum generation cost in a microgrid controlled unit as the target. In which the different confidence level are considered in the expression of the objective function and constraint conditions of rotation backup, and the randomness of the system is reflected by the different confidence levels. A optimization of the established model is performed by using the Monte Carlo stochastic simulation and particle swarm optimization algorithm. What the model obtains is the upper limit of the actual cost with the confidence level not lower than a given value of controllable unit, and the result is even closer to the actual output of the micro grid.5.In order to consider the performance index of the system such as generating capacity, production cost and reliability with the microgrid unit operating in the optimal output way within a period of time in the future, a stochastic production simulation of the micro grid is made. The Time of the photovoltaic output in the micro grid system is divided into two hour intervals, namely, day and night, and the output is used as a negative load; The output of wind power is modeled into multiple state unit output model, and a stochastic simulation and analysis of the output of generating units is conducted with the period as 1 year (8760 h), half a year (4368 h) and a month(720 h) of 30 days by equivalent-power method and semi-invariable method.