风电短期功率预测与并网多目标调度优化研究

发布时间：2018-03-28 17:00

本文选题：风功率预测　切入点：集合经验模态分解　出处：《武汉大学》2014年博士论文

【摘要】：风电具有较强的间歇性和波动性,风电并网容量在电力系统中比重的增加,会给电网带来一些不利影响,一定程度上影响电网的电能质量,如电压暂降和暂升、谐波问题的频率偏差等,同时会降低系统调度运行的可靠性,增加系统调度运行成本。而提高风功率预测水平和建立并求解合理的风电并网情况下电力系统调度模型都能提高风电利用水平,降低风电并网后对电网造成的冲击。在此背景下,本文基于风功率预测和风电并网后电力系统多目标有功调度优化展开研究,所做的主要工作如下： (1)从研究风功率特性出发,提出了一种基于集合经验模态分解和小波神经网络的新型风功率组合预测模型。首先使用集合经验模态分解降低风功率信号的非平稳性,接着采用相空间重构挖掘各子序列信号的混沌特性,然后运用小波神经网络对各子序列分别进行建模预测,最后将各子序列的预测结果叠加得到最终预测结果。实例研究表明,本文所提的组合预测模型具有较高的预测精度和较大的工程应用潜力。 (2)针对传统建模流程忽视预测模型学习效果这一弊端,结合自适应扰动粒子群算法和学习效果反馈机制,建立了一种基于集合经验模态分解和最小二乘支持向量机的风速组合预测模型。首先使用集合经验模态分解将风速信号中真实存在的不同尺度趋势或波动逐级分解出来,然后对分解得到的子序列分别构建最小二乘支持向量机预测模型,并基于自适应扰动粒子群算法和模型学习效果反馈机制综合优化各模型参数,最后将各子序列预测结果叠加得到风速预测值。实例研究表明,本文所提的组合预测方法取得了较好的预测效果。 (3)提出了一种精英云变异多目标反向学习粒子群算法来求解计及风电并网的电力系统多目标调度。通过引入备用选择集获得历史帕累托最优解集和全局帕累托最优解集,来提高算法的全局搜索能力；同时在粒子群算法中引入精英云变异和反向学习来提高算法性能,减小算法陷入局部最优的概率；然后通过微距扰动操作来寻找更真实的Pareto前沿；最后基于模糊理论选择折中最优解。将本文算法应用于IEEE6机30节点的调度测试系统,仿真结果验证了本文算法的有效性和先进性。同时本文的研究结果认为在对风电并网容量进行规划时,除了应考虑环境、经济等因素外,更需要同时兼顾电力系统的安全稳定性。 (4)电力系统动态环境经济调度在考虑各个时间断面之间的联系的情况下,以同时优化常规机组的燃料成本最少和污染气体排放量最小为目标,更符合电力系统调度运行实际。然而目前鲜有文献研究计及风电并网情况下的电力系统动态环境经济调度问题。在此背景下,本文在建立计及风电并网后的电力系统动态环境经济调度模型的基础上,探索采用一种多目标量子人工鱼群算法进行模型求解,以期实现风电并网后电力系统调度运行中的环境经济性能指标综合最优。 (5)提出了一种基于精英云变异多目标反向学习粒子群算法和多目标量子人工鱼群算法的混合Pareto最优解集法。基本思想是首先将两种算法并行优化,各自得到最优解集；其次将各自得到的最优解集进行混合,然后对最优解集进行有效修整,从而得到新的Pareto最优解集。将上述算法应用于优化G省电网节能调度模型中,取得了较好的优化效果。与原有调度方案对比后发现,采用混合Pareto最优解集法优化后的调度方案在实验当天比原有调度方案节约购电成本55535.91元,减少二氧化碳排放2557.9533吨,减少二氧化硫22.30368吨,创造了较好的经济和社会效益。
[Abstract]:Wind power has strong intermittent and wave action , and the proportion of wind power grid capacity in electric power system is increased , which affects the power quality of power grid , such as voltage sag and transient rise , frequency deviation of harmonic problem , etc .

( 1 ) Based on the study of wind power characteristics , a new model of wind power combined prediction based on ensemble empirical mode decomposition and wavelet neural network is presented . Firstly , the non - stationarity of the wind power signal is reduced by using a set of empirical mode decomposition , then the chaotic characteristics of each sub - sequence signal are calculated by using wavelet neural network . Finally , the prediction result of each sub - sequence is superimposed to obtain the final prediction result .

( 2 ) In view of the disadvantages of the traditional modeling process that neglect the learning effect of the prediction model , a combined empirical mode decomposition and a learning effect feedback mechanism are combined to establish a combined empirical mode decomposition and least square support vector machine based wind speed combination forecasting model .

( 3 ) An elite cloud variation multi - objective inverse learning particle swarm optimization algorithm is proposed to solve the multi - objective scheduling of power system .
At the same time , we introduce elitist cloud variation and reverse learning in particle swarm optimization algorithm to improve the performance of the algorithm and reduce the probability that the algorithm gets into local optimum .
then finding a more realistic Pareto front edge through the micro - distance perturbation operation ;
In the end , based on fuzzy theory , the optimal solution is selected . The algorithm is applied to the scheduling test system of IEEE 6 - 30 node . The simulation results verify the validity and advancement of the algorithm . At the same time , the results of this paper consider that in planning the wind power grid capacity , besides the environmental and economic factors , the safety and stability of the power system should be taken into account .

( 4 ) The dynamic environmental economic dispatch of power system is based on the consideration of the connection between different sections of time , so as to optimize the fuel cost of the conventional unit and the minimum pollution gas discharge , which is more consistent with the operation of power system dispatching . However , there are few literatures and the dynamic environment economic scheduling problem of power system in the case of wind power grid network . In this background , this paper explores the model solving by using a multi - target quantum artificial fish swarm algorithm , in order to realize the comprehensive optimization of environmental economic performance indexes in the operation of power system dispatching after wind power grid .

( 5 ) A hybrid Pareto optimal solution method based on multi - objective inverse learning particle swarm optimization algorithm and multi - target quantum artificial fish swarm optimization algorithm is proposed .
A new Pareto optimal solution set is obtained by mixing the obtained optimal solution sets , and then the optimal solution set is effectively trimmed to obtain a new Pareto optimal solution set . Compared with the original scheduling scheme , the optimized scheduling scheme is applied to the optimization of the energy saving scheduling model of the G power grid , and the optimal solution set is obtained . Compared with the original scheduling scheme , the optimized scheduling scheme is saved by 555.91 yuan compared with the original scheduling scheme , the carbon dioxide emission is reduced by 2557.9533 tons , the sulfur dioxide is reduced by 22.30368 tons , and better economic and social benefits are created .

【学位授予单位】：武汉大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TM614;TM73

【相似文献】