| At this stage,the global economic development has a sharp increase in demand for electricity.Due to the limited natural fossil energy reserves in the world and the increased awareness of environmental protection,renewable energy power generation technologies such as wind power and solar energy have gradually become hot spots in the energy industry.However,wind power and photovoltaic power generation are uncontrollable,and large-scale energy grids pose severe challenges to the safety and dispatch of power systems.Therefore,it is particularly important to study the prediction accuracy of wind power and the scheduling of wind power and photovoltaic power generation systems.(1)Based on the analysis of domestic and foreign existing cleaning technologies for wind power forecasting SCADA data processing problems in the early stage,this paper proposes a new data cleaning method based on the actual reference power curve.Optimal data cleaning method to use the group variance,respectively,the maximum and minimum region as a reference point,the reference point is offset from the data as abnormal data.The most deadly question of this method is how to determine the threshold(offset).Thresholds are generally determined by artificial repeated trials,but this method of threshold determination is less efficient and not universal.For these problems,given the following solutions:First,according to the actual operating data of the wind turbine farm,the actual power curve fitted using a method Bin.Then,based on the power curve,data cleaning is performed using a two-way intra-group variance method.The improved method not only greatly reduces the dependence on the threshold in the process of data identification,but also realizes the intelligent identification of abnormal data,avoids manual repeated trials,and reduces the workload.Finally,combined with the actual operation data of the Dazhou Chong wind farm,the effectiveness of the improved cleaning method was verified.(2)Based on the data cleaning technology,a short-term power forecasting method combining wind forecasting model and power forecasting model was proposed.This part of the work is divided into two steps:First,take the wind speed series in the original data set as the research object and build a wind speed forecasting model.Then,after cleaning the data as a research object,a power prediction model was constructed.The advantages of step-by-step are:to reduce the interference of "dirty data" on the power prediction model to build a high-precision mapping model.In addition,the principal component analysis method is used to select the model input items,and the support vector machine model parameters are optimized by three methods:grid method,genetic algorithm and particle swarm algorithm.Finally,the simulation and analysis of the experiment are carried out based on the SCADA data of the Dazhou Chong wind farm in Hunan.(3)This paper applies the short-term power prediction results to the source-slave scheduling system.In view of the mismatched usable power and system natural load,an integrated dispatching system is put forward in this paper,which integrates the generation-side resources such as wind power,PV,thermal power and the demand-side resources such as electric vehicles and adjustable load.On the one hand,based on the fuzzy chance constrained programming model,the non-deterministic problems are equivalently rewritten for problem modeling and solving.On the other hand,considering the cost of electricity generation and consumption in the optimization objectives,the generation-side and demand-side interaction mechanism was applied in the constraints,so as to improve the supply-and-demand relationship between the renewable energy's output and load at a lower cost.Influences of the participation degrees of electric vehicles and demand response on the dispatching results were tested by digital simulation in this paper.The results showed that the proposed method can not only match the load and output,but also effectively raise the economic efficiency. |
PDF Full Text Request