• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文学位 >Neural network approaches to power system short-term load forecasting.
【24h】

Neural network approaches to power system short-term load forecasting.

机译：神经网络方法用于电力系统短期负荷预测。

获取原文
获取原文并翻译 | 示例
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

System load forecasting plays one of the most critical roles in modern power system control centers. The accuracy of forecasting influences the decision-making in unit commitment, hydro-thermal coordination, fuel allocation, and off-line network analysis. Accurate system load forecasting is a potential source of great savings for electric utilities. Since the mid-sixties, much research has been devoted to the development of accurate and efficient load forecasting methods. Many approaches used in time series prediction have been applied to power system load forecasting, such as linear regression, exponential smoothing, stochastic process, and state space methods.; Some important issues, however, remain in this field. It is difficult to model the relationships between system load and factors that influence it. Another difficulty lies in estimating and adjusting the model parameters. These parameters are estimated from historical data and they may become obsolete or may not be able to reflect load pattern changes for short time periods. In addition, load forecasting models rely heavily on the particular utility environment in which the models are developed, therefore, they are not sufficiently general to be transferred easily from one company to another.; Neural network theory is a potentially powerful technique for solving the above problems. The study presented in this dissertation is devoted to the development of system load forecasting methods based on neural networks. Two types of neural network models, i.e., feedforward and adaptive, are investigated for this application. The concepts and methodology formulated have been tested by using load and weather data from power utilities. The testing shows that the proposed approaches have superior qualities in dealing with the difficulties encountered in the traditional techniques. In addition, the numerical simulation results show that this new approach may provide more accurate forecasts than the traditional methods.; This dissertation also includes a comparison of the neural network approach and other established techniques in this field. It is demonstrated that the former provides a parallel representation of the latter. The algorithms and concepts developed in the neural network research can be used to improve the existing techniques.; The new neural network technique is highly feasible as a practical load forecasting tool. As a research project, a prototype neural network model is under development for the power control center at the Pacific Gas and Electric Company. The author has been carrying out this project at the company since the summer of 1991. The design concepts and initial testing results are presented in this dissertation. The results indicate that the prototype model is competitive with the existing forecasting program in the control center.
机译：系统负载预测在现代电力系统控制中心中扮演着最关键的角色之一。预测的准确性会影响机组承诺，水热协调，燃料分配和离线网络分析的决策。准确的系统负载预测是电力企业节省大量资金的潜在来​​源。从六十年代中期开始，许多研究致力于精确和有效的负荷预测方法的发展。时间序列预测中使用的许多方法已应用于电力系统负荷预测，例如线性回归，指数平滑，随机过程和状态空间方法。但是，该领域中仍然存在一些重要问题。很难对系统负载和影响它的因素之间的关系进行建模。另一个困难在于估计和调整模型参数。这些参数是根据历史数据估算得出的，它们可能会过时或可能无法在短时间内反映负载模式的变化。此外，负荷预测模型在很大程度上依赖于开发模型的特定公用事业环境，因此，它们的通用性不足以轻松地从一家公司转移到另一家公司。神经网络理论是解决上述问题的潜在强大技术。本文的研究致力于基于神经网络的系统负荷预测方法的发展。为此应用研究了两种类型的神经网络模型，即前馈和自适应。通过使用来自电力公司的负荷和天气数据，对制定的概念和方法进行了测试。测试表明，所提出的方法在处理传统技术中遇到的困难方面具有卓越的品质。另外，数值模拟结果表明，这种新方法可能比传统方法提供更准确的预测。本文还对神经网络方法与该领域其他已建立的技术进行了比较。事实证明，前者提供了后者的并行表示。神经网络研究中开发的算法和概念可用于改进现有技术。新的神经网络技术作为一种实用的负荷预测工具是高度可行的。作为一个研究项目，正在为太平洋天然气和电力公司的功率控制中心开发原型神经网络模型。自1991年夏季以来，作者一直在公司中执行此项目。本文介绍了设计概念和初步测试结果。结果表明，该原型模型与控制中心现有的预测程序相比具有竞争力。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号