微网功率与负荷的不确定预测及经济调度

发布时间：2019-02-09 13:38

【摘要】：由于微网系统中的分布式发电和负荷均具有间歇性、随机性、波动性等多种不确定性,如何优化调度各种分布式发电以保证微网系统的供电可靠性和经济运行性,成为当前微网研究领域的热点之一。对含不确定性的可再生能源出力和负荷进行准确、可靠的预测,进而构建并求解含上述不确定性的微网经济调度模型是解决上述问题的关键。鉴于此,针对微网系统中可再生能源出力和负荷的不确定性预测及经济调度问题进行了深入研究,主要研究内容如下:(1)基于回声状态网络的风电功率不确定性区间预测:考虑风力发电所具有的极大的不确定性,针对当前大多基于点预测而不能反应数据本质不确定性的不足,同时为充分利用回声状态网络在时序预测中的优势,提出一种基于回声状态网络的定量化描述不确定性的风电功率区间预测模型。首先,利用相似日理论和区间相似准则,给出新的体现风电功率不确定性的区间样本选择策略;然后,基于样本的区间表达,给出基于回声状态网络的风电功率时序区间预测方法;仿真实验通过区间覆盖率、区间平均宽度等指标验证了所提方法的有效性。(2)基于集成极限学习机的负荷不确定性区间预测:首先分析不同类型的微网负荷所具有的强波动不确定特性;然后给出基于统计分析的负荷波动关键影响因素的提取策略,由于传统的基于相似日数据的全局逼近难以保证拟合的精度且无法体现不确定性,进一步进行分时段相关性分析,提出日期类型-温度相似性的相似日度量准则,以选择合适的样本;基于多相似日数据和极限学习机构造集成预测模型,并利用Bootstrap抽样方法给出集成预测结果的置信区间,从而获得微网负荷的不确定性预测区间,仿真实验验证了所提方法的有效性。(3)基于NSGA-II的含不确定性的微网多目标经济调度:在上述研究内容的基础上,结合储能装置实时容量、实时电价等影响因素,建立含风电等分布式发电和用电负荷的不确定性微网系统多目标经济调度模型;基于对风电和负荷的不确定性区间预测结果,给出分时段的模型优化策略;在风力发电不满足负荷需求时,给出了基于多目标遗传算法NSGA-II的调度优化求解结果;仿真实验深入分析了风电出力和负荷波动对调度计划的影响,表明了考虑风电和负荷不确定性的微网经济调度的合理性。本文提出的预测方法,能较好的反映风电功率和负荷预测可能的变化范围,考虑不确定性得到的优化结果比不计不确定性的方法会具有更好的适应性和鲁棒性,为实际应用提供了较为可靠的理论参考。
[Abstract]:Because the distributed generation and load in microgrid system have many uncertainties, such as intermittence, randomness, volatility and so on, how to optimize and dispatch all kinds of distributed generation to ensure the reliability and economic operation of micro-grid system. It has become one of the hotspots in the field of microgrid research. The key to solve the above problems is to accurately and reliably predict the output and load of renewable energy with uncertainty, and then construct and solve the micro-grid economic scheduling model with the uncertainty mentioned above. In view of this, the uncertain forecasting and economic scheduling problems of renewable energy output and load in microgrid systems are studied. The main research contents are as follows: (1) interval prediction of wind power uncertainty based on echo state network: considering the great uncertainty of wind power generation. In order to make full use of the advantage of echo state network in time series prediction, most of the shortcomings of point prediction can not reflect the essential uncertainty of data at present, so as to make full use of the advantage of echo state network in time series prediction. An interval prediction model of wind power based on echo state network is proposed to quantitatively describe uncertainty. Firstly, by using the similarity day theory and the interval similarity criterion, a new interval sample selection strategy for wind power uncertainty is proposed. Then, based on the interval representation of samples, an interval prediction method of wind power time series based on echo state network is presented. The simulation experiment passes through the interval coverage, The effectiveness of the proposed method is verified by interval average width and other indicators. (2) interval prediction of load uncertainty based on integrated LLMs: firstly, the strong volatility uncertainty characteristics of different types of microgrid loads are analyzed. Then the extraction strategy of the key factors of load fluctuation based on statistical analysis is given. Because the traditional global approximation based on similar day data can not guarantee the precision of fitting and can not reflect the uncertainty, the correlation analysis is further carried out. The similarity day measure criterion of date type-temperature similarity is proposed to select suitable samples. An integrated prediction model is constructed based on multi-similar daily data and extreme learning machine, and the confidence interval of integrated prediction results is given by using Bootstrap sampling method, and the uncertainty prediction interval of micro-grid load is obtained. The simulation results show that the proposed method is effective. (3) Multi-objective economic scheduling with uncertainty based on NSGA-II: based on the above research, combined with the real time capacity of energy storage device, real time electricity price and other influencing factors, A multi-objective economic scheduling model for uncertain micro-grid systems with distributed generation and load, such as wind power, is established. Based on the uncertain interval prediction results of wind power and load, the model optimization strategy is proposed, and the scheduling optimization results based on multi-objective genetic algorithm (NSGA-II) are given when wind power generation does not meet the load demand. The effect of wind power output and load fluctuation on scheduling plan is analyzed in detail by simulation experiments. The rationality of micro-grid economic scheduling considering wind power and load uncertainty is demonstrated. The prediction method presented in this paper can better reflect the possible range of wind power and load forecasting. The optimization results obtained by considering uncertainty have better adaptability and robustness than those without uncertainty. It provides a reliable theoretical reference for practical application.
【学位授予单位】：中国矿业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM73

【相似文献】