基于实测数据的大规模光伏出力特性及其短期预测方法研究

发布时间：2018-10-15 08:18

【摘要】：光伏电站群集中并网发电是太阳能大规模开发利用的重要途径,随着光伏电站装机容量的不断增加,高渗透率光伏功率的波动将会对电网造成一系列消极的影响。对光伏电站以及光伏电站群出力波动特性的全面分析以及光伏功率的准确预测是研究光伏并网相关问题的基础。本文基于青海省大型光伏发电基地的实测数据,首先对光伏电站功率特性及其汇聚后电站群功率特性进行了分析。研究了单一光伏电站的日出力特性以及天气与季节变化对输出功率的影响。构建波动特性指标,分析了不同时间尺度以及不同装机容量下光伏功率波动特性。从光伏电站间功率相关性的角度,揭示了光伏电站群的汇聚效应,提出汇聚系数概念衡量光伏电站群的汇聚效应。指出汇聚效应的应用方向并提出计及汇聚效应的光伏电站群输出功率预测思路。介绍灰色神经网络模型应用到光伏预测的预测原理,并对其运用到光伏预测时的适用性进行分析,根据分析结果对原始功率序列进行平滑处理以改进灰色模型,针对BP神经网络的缺陷采用粒子群算法对其进行优化,构建改进灰色神经网络组合模型实现对单一光伏电站提前一天的短期功率预测。算例结果表明,改进后模型的预测精度较改进前的灰色神经网络模型有明显提高,并且满足国家能源局设置的预测误差标准。基于光伏电站群的相关性分析,提出一种计及汇聚效应的区域光伏电站群短期功率预测方法。该方法根据相关性计算结果选出基准光伏电站并对其进行预测,对预测值进行线性放大得出光伏电站群预测的估计值,最后根据各光伏电站间的相关系数对估计值进行修正,实现对光伏电站群的短期功率预测。算例结果表明,与常用的叠加法相比,该方法的预测结果更接近实际值并且预测精度有明显的提高,并且区域光伏电站群的预测精度要高于单一光伏电站的预测精度。
[Abstract]:Centralized grid-connected photovoltaic power generation is an important way of large-scale development and utilization of solar energy. With the increasing of installed capacity of photovoltaic power station, the fluctuation of photovoltaic power with high permeability will cause a series of negative effects on power grid. The comprehensive analysis of the fluctuation characteristics of the output force of photovoltaic power stations and the accurate prediction of photovoltaic power are the basis of the research on grid-connected photovoltaic problems. Based on the measured data of large photovoltaic power station in Qinghai province, the power characteristics of photovoltaic power station and the power characteristics of converged power station group are analyzed in this paper. The daily output characteristics of a single photovoltaic power plant and the effects of weather and seasonal variations on the output power are studied. The fluctuation characteristics of photovoltaic power are analyzed under different time scales and different installed capacity. From the point of view of power correlation between photovoltaic power stations, the convergent effect of photovoltaic power station group is revealed, and the concept of convergence coefficient is proposed to measure the convergence effect of photovoltaic power plant group. The application direction of convergent effect is pointed out, and the forecast thought of output power of PV power station group considering convergent effect is put forward. This paper introduces the prediction principle of the grey neural network model applied to the photovoltaic prediction, and analyzes its applicability when it is applied to the photovoltaic prediction. According to the analysis results, the original power series is smoothed to improve the grey model. Particle swarm optimization (PSO) is used to optimize the BP neural network, and an improved grey neural network combination model is constructed to predict the short term power of a single photovoltaic power station one day in advance. The calculation results show that the prediction accuracy of the improved model is obviously higher than that of the grey neural network model before the improvement, and the prediction error standard set by the State Energy Bureau is satisfied. Based on the correlation analysis of photovoltaic power station group, a short-term power prediction method for regional photovoltaic power station group is proposed, which takes into account the convergent effect. The method selects and forecasts the reference photovoltaic power station according to the results of correlation calculation. The predicted value is obtained by linear amplification of the predicted value. Finally, the estimated value is revised according to the correlation coefficient among the photovoltaic power stations. The short-term power prediction of PV power station group is realized. The numerical results show that the prediction results of the proposed method are closer to the actual values and the prediction accuracy of the regional PV power stations is higher than that of the single photovoltaic power station compared with the common superposition method.
【学位授予单位】：东北电力大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM615

【参考文献】