分布式光伏发电功率预测及优化配置研究

发布时间：2018-10-20 15:00

【摘要】：随着国际社会对能源、生态、气候等问题的重视,大力发展可再生能源成为应对这些问题的重要举措。太阳能作为可再生能源中的一员,近年来发展迅速,分布式光伏作为光伏产业下一步大力发展的对象,具有运用方式灵活,节约电能损耗,能源利用率高等特点,是智能电网配电环节的重要组成部分。由于光伏输出呈现间歇性、不稳定的特点,随着分布式光伏在配网中渗透率的增加,配电网的结构与运行方式将会发生变化,为了减少对配网的影响,需要对分布式光伏进行短期功率预测,而且还要对分布式光伏在配网中的位置以及容量进行优化。因此,本文搭建了分布式光伏短期功率预测模型,对收集到的样本进行预测并对结果作误差分析;在分布式光伏优化配置方面,建立了多目标优化模型,运用改进多目标微分进化算法,对标准配网中的分布式光伏进行优化配置,主要研究内容如下:(1)研究了光伏发电系统的组成与分类,通过对理论模型与实际数据的仿真,从影响光伏功率输出的角度,研究了光伏发电功率特性,并确定出光伏预测模型的输入。从配电网系统潮流、供电可靠性、电能质量等方面分析了分布式光伏并网对配电网所造成的影响,并从中选取多目标优化模型的优化目标与约束条件。(2)研究了小波变换的基本原理,对采集到的光伏功率序列作小波变换预处理,得到功率的各层序分量,分别建立预测模型进行预测。分析了ESN神经网络的原理与结构,明确了ESN网络训练的方法。针对收集到的光伏数据确定出预测模型的结构,输入以及输出参数,搭建了WT+ESN光伏功率预测模型。(3)在matlab平台上,选取晴天样本与阴天样本,分别运用ESN,BP,WT+ESN以及WT+BP四种预测模型对分布式光伏的出力进行预测,并运用三种误差指标对预测结果进行评估。仿真结果表明,WT+ESN相较于其它三种模型,预测曲线更平稳,变化趋势更贴近实际曲线,三种误差评价指标均是最优的,验证了WT+ESN预测模型的有效性与优越性。(4)建立了包含分布式光伏成本与运维费用,配电网有功网损,电压稳定指标VSI以及各种约束条件的多目标优化模型。介绍了多目标优化问题的数学描述。研究了微分进化算法的原理与算法流程,鉴于传统微分进化算法过度依赖经验控制参数,将自适应策略融入到算法中,并结合Pareto占优概念提出了改进多目标微分进化算法—MOSADE。(5)针对IEEE-33节点的标准配网,在matlab平台上进行分布式光伏优化配置研究。仿真结果表明,在网损最优方面,SADE与DE及LDWPSO算法相比,在算法前期保持了种群多样性,在算法后期提升了收敛速度;在全局最优方面,运用MOSADE对1-10组的DPV进行有效配置,结果表明DPV的合理配置可以有效提升配网电压水平、减小有功网损、增加发电收益,验证了多目标优化模型以及MOSADE算法的合理性与有效性。
[Abstract]:With the attention of the international community on energy, ecology, climate and other issues, vigorously developing renewable energy has become an important measure to deal with these problems. Solar energy, as a member of renewable energy, has developed rapidly in recent years. Distributed photovoltaic, as the next step of the photovoltaic industry, has the characteristics of flexible application, energy saving, high energy utilization and so on. It is an important part of the distribution link of smart grid. Because of the intermittent and unstable characteristics of photovoltaic output, with the increase of permeability of distributed photovoltaic in distribution network, the structure and operation mode of distribution network will change, in order to reduce the influence on distribution network. It is necessary to predict the short term power of distributed photovoltaic, and to optimize the position and capacity of distributed photovoltaic in distribution network. Therefore, this paper builds a distributed PV short-term power prediction model, forecasts the collected samples and makes error analysis of the results, and establishes a multi-objective optimization model for the distributed photovoltaic optimal configuration. Using the improved multi-objective differential evolution algorithm, the distributed photovoltaic system in standard distribution network is optimized. The main contents are as follows: (1) the composition and classification of photovoltaic power generation system are studied, and the simulation of theoretical model and practical data is carried out. From the point of view of affecting the output of photovoltaic power, the characteristics of photovoltaic power generation are studied, and the input of photovoltaic prediction model is determined. The influence of distributed photovoltaic grid connection on distribution network is analyzed from power flow, power supply reliability, power quality and so on. And the optimization objectives and constraints of the multi-objective optimization model are selected. (2) the basic principle of wavelet transform is studied. The collected photovoltaic power sequence is preprocessed by wavelet transform to obtain the sequence components of the power. The prediction models are established respectively. The principle and structure of ESN neural network are analyzed, and the training method of ESN neural network is clarified. According to the collected photovoltaic data, the structure, input and output parameters of the prediction model are determined, and the WT ESN photovoltaic power prediction model is built. (3) on the matlab platform, the sunny and overcast samples are selected. Four prediction models, ESN,BP,WT ESN and WT BP, are used to predict the output of distributed photovoltaic, and three kinds of error indexes are used to evaluate the prediction results. The simulation results show that compared with the other three models, the prediction curve of, WT ESN is more stable, the trend of change is closer to the actual curve, and the three kinds of error evaluation indexes are all the best. The validity and superiority of WT ESN prediction model are verified. (4) A multi-objective optimization model including distributed photovoltaic cost and operation cost, distribution network active power loss, voltage stability index (VSI) and various constraints is established. The mathematical description of multi-objective optimization problem is introduced. In this paper, the principle and flow of differential evolution algorithm are studied. In view of the traditional differential evolution algorithm relying too much on empirical control parameters, adaptive strategy is integrated into the algorithm. Combined with the concept of Pareto dominance, an improved multi-objective differential evolutionary algorithm (MOSADE. (5) is proposed for the standard distribution of IEEE-33 nodes. The distributed photovoltaic optimal configuration is studied on the matlab platform. The simulation results show that, compared with DE and LDWPSO algorithms, SADE maintains population diversity in the early stage of the algorithm, and improves the convergence speed in the later stage of the algorithm. In the aspect of global optimization, MOSADE is used to effectively configure 1-10 groups of DPV. The results show that the reasonable allocation of DPV can effectively enhance the voltage level of distribution network, reduce the loss of active power network, increase the income of generation, and verify the rationality and validity of the multi-objective optimization model and MOSADE algorithm.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM615

【参考文献】