基于数据挖掘和GSA-BP多模型神经网络的微网短期负荷预测

发布时间：2018-11-10 19:08

【摘要】：随着我国经济的快速发展,能源枯竭问题日益加重,微网作为一种清洁友好型能源,能够利用当地的资源优势,有效解决我国中西部偏远地区用电难、电能运输成本高、利用率低等问题。而微网短期负荷预测作为微网中的研究热点,日益受到研究人员的关注。对微网短期负荷进行有效预测,能够为微网系统节能高效的运行提供保障、为电力调度部门制定发电计划提供依据。因此加强微网的负荷预测无论对微网系统本身还是对大电网都有重要的意义。本文针对微网负荷的特点,提出了基于数据挖掘和遗传模拟退火算法(GSA算法)优化的多模型神经网络微网短期负荷预测模型。主要研究工作以及创新性内容如下:首先,对影响微网负荷的气象、日类型以及实际历史负荷等因素进行分析,根据这些因素建立微网负荷预测的初步样本数据,利用数据挖掘技术对该样本数据进行挖掘处理,并建立基本预测模型。具体处理方法为:(1)利用粗糙集属性约简算法对样本数据进行约简处理,找到影响微网负荷的核心因素,并将其作为预测模型的输入;(2)针对微网负荷的波动性、随机性等特点,对约简后的样本数据进行模糊聚类分析,将样本数据分为若干类,并针对每一类样本建立对应的BP神经网络预测模型;(3)在对微网负荷进行预测时,通过模式识别技术寻找与预测日最近的样本集合对应的网络,并利用此网络对微网负荷进行预测。通过以上几步建立了基于多模型BP网络的微网短期负荷预测模型。通过仿真对比,验证了预测模型能够取得理想的预测结果。其次,针对BP神经网络迭代速度慢、易陷入局部极小值等缺陷,提出了基于GSA算法优化的多模型BP网络预测模型,将遗传算法(GA)的并行搜索结构和模拟退火算法(SA)的概率突跳特性与BP网络相结合,对微网负荷做出预测。将优化后的预测模型与优化前的预测模型做对比,结果表明,优化后的模型预测精度更高。并且通过与其他预测算法比较,进一步验证了GSA算法优化的多模型BP网络在微网短期负荷预测中的优势。最后,通过分析国外微网负荷的运行情况发现,实时电价因素在一定程度上会影响微网负荷的大小。因此,本文将实时电价因素引入预测模型中。利用模糊控制算法对模型预测后的微网负荷值进行修正。仿真结果表明,该算法能够对考虑实时电价因素的预测结果进行有效地修正。
[Abstract]:With the rapid development of economy in our country, the problem of energy depletion is getting worse and worse. As a kind of clean and friendly energy, microgrid can effectively solve the problem of electricity utilization in remote areas of central and western China, and the cost of electricity transportation is high. Low utilization rate and other problems. Microgrid short-term load forecasting, as a research hotspot in micro-grid, has been paid more and more attention by researchers. The short-term load forecasting of microgrid can provide the guarantee for the energy saving and efficient operation of the micro-grid system, and provide the basis for the power dispatching department to formulate the generation plan. Therefore, it is of great significance to strengthen the load forecasting of microgrid for both microgrid system and large power grid. According to the characteristics of microgrid load, a multi-model neural network short-term load forecasting model based on data mining and genetic simulated annealing algorithm (GSA) is proposed in this paper. The main research work and innovative contents are as follows: firstly, the factors affecting microgrid load, such as meteorology, daily type and actual historical load, are analyzed, and the preliminary sample data of microgrid load forecasting are established according to these factors. The data mining technology is used to mine the sample data, and the basic prediction model is established. The specific processing methods are as follows: (1) the rough set attribute reduction algorithm is used to reduce the sample data to find the core factors that affect the load of the microgrid and take it as the input of the prediction model; (2) according to the characteristics of volatility and randomness of microgrid load, the reduced sample data is analyzed by fuzzy clustering, and the sample data is divided into several categories, and the corresponding BP neural network prediction model is established for each kind of sample. (3) in forecasting the load of microgrid, the pattern recognition technique is used to find the network corresponding to the nearest sample set on the forecasting day, and the network is used to forecast the load of microgrid. The short term load forecasting model of micro grid based on multi-model BP network is established through the above steps. The simulation results show that the prediction model can obtain ideal prediction results. Secondly, aiming at the shortcomings of BP neural network, such as slow iterative speed and easy to fall into local minimum, a multi-model BP network prediction model based on GSA algorithm optimization is proposed. The parallel search structure of genetic algorithm (GA) and the probabilistic jump characteristics of simulated annealing algorithm (SA) are combined with BP network to predict the load of microgrid. The results show that the precision of the optimized model is higher than that of the optimized model. Compared with other prediction algorithms, the advantages of multi-model BP network optimized by GSA algorithm in short-term load forecasting of microgrid are further verified. Finally, by analyzing the operation of microgrid load in foreign countries, it is found that the real time price factor will affect the load of microgrid to some extent. Therefore, this paper introduces the real-time electricity price factor into the prediction model. The fuzzy control algorithm is used to modify the load value of the microgrid after the model prediction. The simulation results show that the algorithm can effectively modify the prediction results considering the real-time price factors.
【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】：TP311.13;TP18;TM715

【参考文献】