基于神经网络预测控制的微电网能量优化管理研究

发布时间：2018-04-15 01:24

本文选题：预测控制 + GA-BP神经网络　；参考：《湖南工业大学》2017年硕士论文

【摘要】：随着全球能源和环境问题的日益突出,全世界各国都开始重视能源危机、环境破坏等问题,并着力于寻找清洁能源补充和改善当下境况。近些年来,我国利用可再生能源的比重逐年增加,分布式发电项目在海岛和偏远地区也不断投产建设。此外,由分布式发电单元组成的微电网系统与网络控制系统相结合组成的智能微电网,也大大降低了微电网系统的运营成本,提高了微电网的稳定性、灵活性,扩宽了其应用领域。然而,基于网络控制的微电网系统也存在不足之处,比如分布式单元的独立稳定运行、各分布式单元的能量切换、微电网的低压穿越等问题。如果这些问题在微电网实际运行过程中得不到有效地解决,那么微电网系统将陷入不稳定运行或者瘫痪,同时也会给用户侧用电设备造成伤害。另外,网络控制是通过周期性地检测系统信号,然后将数据传送至中央控制器进行控制决策,由于网络通信固有的缺陷,如通信丢包、迟滞等,难免会出现通信失误,并且微电网主要分布式发电单元受气象因素的影响,其波动性和随机性也使得发电功率输出具有不确定性,故在精准控制上需要引入新的控制方法。鉴于此,本文分析国内外许多专家和学者在微电网结构和控制策略的相关文献,提出了基于神经网络预测控制的微电网能量优化管理,文中在已有基础上加入了预测控制,为微电网的合理调度和控制决策提供了参考值;文中还引入了新的微电网能量管理结构,使得资源配置更加合理,能量管理更加优化。最后,具体的创新性研究工作从以下三个方面介绍:(1)预测控制模型采用的主预测工具是神经网络,并加入反向传播途径,即为BP神经网络,为了缩短收敛周期,文中还引入遗传算法,但仍然不能完全避免其收敛周期过长,故引入了回归分析预测来填补收敛周期过长的情况,最后文中将神经网络与回归分析相结合构成预测的控制机制实现了预测控制的最优化。(2)每个分布式单元都是采用电力电子接口输出,且输出端都有U、I检测,将检测信息送至控制器反馈调节电力电子器件,实现单分布式单元的相对稳定控制输出。此外,文中还将这些检测信号送至中央控制器,并对分布式单元协调控制,使得微电网能量管理更加合理。(3)文中的微电网能量管理结构是多层次结构,且具体分为上层决策层和下层决策层,上层决策层具有一组固定的参数(参考值)函数,并通过特定的控制策略结构来适应较低的决策水平,根据这一参数,每个较低级别的决策者通过跟踪上层决策者所提供的参考值解决自身的优化问题。结构建模上还考虑了最小碳排放和降低生产成本,提高了市场竞争力,使得本次研究更具工程实践价值。
[Abstract]:With the increasingly prominent global energy and environmental problems, countries all over the world have begun to pay attention to energy crisis, environmental damage and other issues, and focus on finding clean energy to supplement and improve the current situation.In recent years, the proportion of renewable energy used in China has increased year by year, and distributed power generation projects have been put into production in islands and remote areas.In addition, the smart microgrid, which is composed of the microgrid system and the network control system, greatly reduces the operating cost of the micro-grid system, and improves the stability and flexibility of the micro-grid system.It widens its application field.However, the microgrid system based on network control also has some shortcomings, such as the independent and stable operation of the distributed unit, the energy switching of each distributed unit, the low-voltage traversing of the micro-grid, and so on.If these problems can not be effectively solved in the actual operation of microgrid, then the microgrid system will be unstable or paralyzed, and will also cause harm to the consumer side power equipment.In addition, network control detects system signals periodically and then transmits data to the central controller for control decision. Due to the inherent defects of network communication, such as packet loss and delay, communication errors occur inevitably.And the main distributed generation units of microgrid are affected by meteorological factors, its volatility and randomness also make the generation power output uncertain, so it is necessary to introduce a new control method in precise control.In view of this, this paper analyzes many domestic and foreign experts and scholars in microgrid structure and control strategy related literature, proposed based on neural network predictive control of microgrid energy optimization management, this paper added predictive control on the basis of the existing.The paper also introduces a new energy management structure of microgrid, which makes resource allocation more reasonable and energy management more optimized.Finally, the specific innovative research work introduces the following three aspects: 1) the main predictive tool used in the predictive control model is the neural network, and the BP neural network is added to the back propagation path. In order to shorten the convergence period, the main predictive tool used in the predictive control model is the BP neural network.Genetic algorithm (GA) is also introduced in this paper, but the convergence period is still too long. Therefore, regression analysis is introduced to replace the case of long convergence cycle.Finally, the neural network and regression analysis are combined to form the predictive control mechanism to realize the optimization of predictive control. (2) every distributed unit is output with power electronic interface, and the output end has UFI detection.The detection information is sent to the controller to feedback and adjust the power electronic device to realize the relatively stable control output of the single distributed cell.In addition, these detection signals are sent to the central controller, and the distributed unit is coordinated to make the energy management of microgrid more reasonable.The upper decision layer has a set of fixed parameters (reference value) function and adapts to the lower decision level through the specific control strategy structure.Each decision maker at the lower level solves the optimization problem by tracking the reference values provided by the upper decision makers.In structural modeling, the minimum carbon emission and production cost are considered, and the market competitiveness is improved, which makes this study more valuable in engineering practice.
【学位授予单位】：湖南工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP183;TM727

【参考文献】